Keywords:
wifi offloading
wifi

Abstract:
... a diverse range of networks; cellular access (3G to LTE), WiFi access (2.4 to 5GHz), deployment of more public WiFi access points (APs), as they use diverse applications such as ... it. We show that users are gradually but steadily adopting WiFi at home, in offices, and public spaces over these three ... majority of light users have been shifting their traffic to WiFi. . Heavy hitters acquire more bandwidth via WiFi, , especially at home. The percentage of users explicitly turning ... at home. The percentage of users explicitly turning off their WiFi interface during the day decreases from 50% to 40%. Our ... improved during the three years, with more than 40% of WiFi users connecting to multiple WiFi APs in one day. WiFi offload at offices is still limited in our dataset due ... in our dataset due to a few accessible APs, but WiFi APs in public spaces have been an alternative to cellular ...

References:
M. Afanasyev, T. Chen, G. M. Voelker, and A. C. Snoeren. Analysis of a mixed-use urban WiFi network: When metropolitan becomes neapolitan. In IMC'08, pages 85--98, Vouliagmeni, Greece, Oct 2008.
A. Balasubramanian, R. Maha jan, and A. Venkataramani. Augmenting mobile 3G using WiFi. In MobiSys'10, pages 209--222, San Francisco, CA, Jun 2010.
X. Chen, R. Jin, K. Suh, B. Wang, and W. Wei. Network performance of smart mobile handhelds in a university campus WiFi network. In IMC'12, pages 315-328, Boston, MA, 2013.
P. Deshpande, X. Hou, and S. R. Das. Performance comparison of 3G and metro-scale WiFi for vehicular network access. In IMC'10, pages 301--307, Melbourne, Australia, Nov 2010.
J. Hare, L. Hartung, and S. Banerjee. Beyond deployments and testbeds: Experiences with public usage on vehicular WiFi hotspots. In MobiSys'12, pages 393--405, Low Wood Bay, UK, Jun 2012.
K. Lee, I. Rhee, J. Lee, S. Chong, and Y. Yi. Mobile data offloading: How much can WiFi deliver? In CoNEXT'10, page 12, Philadelphia, PA, Dec 2010.
A. Patro, S. Govindan, and S. Banerjee. Observing home wireless experience through wifi aps. In MobiCom'13, pages 339--350, Miami, FL, Sep 2013.
J. Sommers and P. Barford. Cell vs. WiFi: On the performance of metro area mobile connections. In IMC'12, pages 301--314, Boston, MA, Nov 2012.

Full Text:
... usersacross a diverse range of networks; cellular access (3G toLTE), WiFi access (2.4 to 5GHz), deployment of more pub-lic WiFi access points (APs), as they use diverse applica-tions such as ... send it. We show that users are gradually but steadilyadopting WiFi at home, in offices, and public spaces overthese three years. ... The majority of light users have beenshifting their traffic to WiFi. . Heavy hitters acquire morebandwidth via WiFi, , especially at home. The percentage ofusers explicitly turning off ... especially at home. The percentage ofusers explicitly turning off their WiFi interface during theday decreases from 50% to 40%. Our results ... been improved during thethree years, with more than 40% of WiFi users connectingto multiple WiFi APs in one day. WiFi o?oad at officesis still limited in our dataset due to ... limited in our dataset due to a few accessible APs,but WiFi APs in public spaces have been an alternative tocellular access ... residential broadband and cellular traf-fic in Japan [34].KeywordsSmartphone traffic; LTE; WiFi; ; WiFi o?oad1. INTRODUCTIONIn recent years the deployment of high-speed Internet athome ... they are currently encouraging theircustomers to o?oad cellular traffic to WiFi networks (WiFio?oading). To enable WiFi o?oading, cellular providers aredeploying new free WiFi access points (APs) for their cus-tomers in public places (e.g., ...
Furthermore, they have begun to give cus-tomers free home WiFi routers with the expectation thatthese will increase o?oading at home ... of supporting Internet access for foreignvisitors. Rapid deployment of better WiFi coverage enablescellular providers to o?oad 3G/LTE traffic to WiFi net-works in many places if they appropriately manage to leadtheir ... manage to leadtheir users to these networks.The success of these WiFi deployment hinges on this keyquestion: how do smartphone users select ... volume accounts for 28% of the totalbroadband traffic volume. (2) WiFi traffic volume has in-creased more than cellular traffic volume due ... of heavy hitters. Even for light users (i.e., me-dian), the WiFi traffic is higher than that of cellular trafficas of 2015 ... 10% in three years ( 3.3). Also, thenumber of associated WiFi networks per user has increased.As of 2015, 40% of WiFi users are associated with at leasttwo APs in one day ... one day ( 3.4). The traffic volume to publicand office WiFi APs, however, only accounts for 2% of thetotal WiFi traffic volume. (4) The signal strength of theassociated WiFi network is generally high; however, 12% ofthe public WiFi networks exhibit subpar quality ( 3.4, and 3.5). We further ... We further confirm the rapid deployment of 5GHzAPs in public WiFi networks though dominant APs at homeand office are still 2.4GHz ... (e.g., video, online storage) have become morepopular ( 3.6) in WiFi networks. We see indications thatyear duration #And #iOS #total %LTE2013 ... and network availability).Our results show slow but clear growth of WiFi traffic of-floading during these three years. In particular, the deploy-ment ... during these three years. In particular, the deploy-ment of public WiFi networks provides users both simplenetwork connectivity and also more bandwidth ... such as video streaming and softwareupdates. However, we find that WiFi o?oading at offices isstill limited.2. METHODOLOGY AND DATASETWe developed measurement ... information, battery status, geolocation information,network information (e.g, BSSID, ESSID for WiFi) ), and aunique random device ID. While most information is ... the traffic volume per application. Similarly, An-droid OS reports non-associated WiFi AP information aswell as associated WiFi APs if the WiFi interface is turnedon, whereas iOS only reports the associated AP ...
... advanced users than the averagesmartphone user. The availability of public WiFi networksin metro stations, cafes, shops, and on the streets is ... andbuses) rather than personal cars. Thus, the probability toencounter public WiFi networks is likely high, and resultingWiFi traffic volume in public ... examine the aggregated traffic behavior of cellular(3G and LTE) and WiFi networks in our datasets. Figure 2indicates the weekly variations in ... figureare traffic volumes from and to smartphones, respectively.As expected, the WiFi volume exceeds the cellular volume.Thus the traditional measurements of cellular ... underrepresent all smartphone activity due to WiFio?oad. The ratio of WiFi traffic to the total traffic increasesover time from 59% in ... day, with cellular traffic peaks corresponding to com-mute times and WiFi peaks to evening times at home. Threetraffic peaks in cellular ... (9pm), though we also confirm peaks for morning andnoon in WiFi RX. Cellular traffic on weekends is smallerthan that on weekdays, ... Cellular traffic on weekends is smallerthan that on weekdays, while WiFi traffic is the opposite.Thus, these observations clearly present different temporalusage ... is the opposite.Thus, these observations clearly present different temporalusage of WiFi and cellular networks.3.2 Daily user traffic volumeWe next focus on ...
... theresults clearly show the client-server type user behavior.Daily cellular and WiFi traffic: Next, we present thatdaily WiFi and cellular traffic volumes per user are largelydistributed. Figure 4 ... the CDFs of the daily usertraffic volume for cellular and WiFi network interfaces in2015. As expected, the download and upload traffic ... skewed. On one hand, 8% of cellular interfacesand 20% of WiFi interfaces do not send and receive anydata. On the other ... 2014 2015 AGRAll 57.9 90.3 126.5 48%Cell 19.5 27.6 35.6 35%WiFi 9.2 24.3 50.7 134%mean 2013 2014 2015 AGRAll 102.9 179.9 ... 2014 2015 AGRAll 102.9 179.9 239.5 53%Cell 42.2 58.5 71.5 30%WiFi 60.7 121.5 168.1 66%Table 3: Daily download traffic volume per ... obtainedby linear fit.We see that users download more traffic via WiFi in lateyears. In fact, the traffic growth of median WiFi RX issignificant while the median cellular RX is higher than ... issignificant while the median cellular RX is higher than themedian WiFi ... RX in 2013. Thus, light users overlook the im-portance of WiFi o?oading in 2013, but they change theirbehavior depending on the ... increase in traffic volume. Sim-ilarly, the growth of the mean WiFi traffic volume is high.Heavy hitters are also heavier in WiFi than cellular in meantraffic volume.3.3 WiFi vs. CellularTo assess the impact of traffic o?oading we next ... traffic o?oading we next exam-ine of how users select between WiFi and cellular networkinterfaces.3.3.1 Aggregated viewWiFi-intensive and cellular-intensive users: We in-troduce ... and cellular-intensive users: We in-troduce three types of users (cellular-intensive, WiFi- -intensive,and mixed user) from the usage of network interfaces. Fig-ure ... 0.8 1Sat Sun Mon Tue Wed Thu Fri SatWiFi-traffic ratio(a) WiFi- -traffic ratio 20132015 0 0.2 0.4 0.6 0.8 1Sat Sun ... 0.8 1Sat Sun Mon Tue Wed Thu Fri SatWiFi-user ratio(b) WiFi- -user ratio 20132015Figure 6: (a) WiFi- -traffic ratio and (b) WiFi- ... -user ratio.each user and each day by network (cellular and WiFi) ). Thecolor bar indicates the number of users per day ... three typical user types; Cellular-intensive users do not use their WiFi interfaces due to noavailable WiFi APs or no WiFi configuration. WiFi- -intensiveusers only rely on their WiFi interfaces to avoid their cel-lular traffic fees. Mixed users select ... figure indicates that 22% of usersare cellular-intensive and 8% are WiFi- -intensive users. Wesee a decrease in cellular-intensive users: from 35% ... 35% in 2013to 22% in 2015, while the fraction of WiFi- -intensive usersis stable during the three years. Interestingly, both WiFi- -intensive and cellular-intensive users show a wide spread intraffic volume; ... wide spread intraffic volume; thus either can be heavy hitters.User-level WiFi o?oading: This data shows that mixed-network users often o?oad traffic ... This data shows that mixed-network users often o?oad traffic to WiFi networks. Usersabove the diagonal are evidence of o?oading, since theirWiFi ...
cellular traffic. The concentrationof users above the diagonal shows WiFi o?oading is com-mon. Although 55% of mixed users are above ... 0.8 1Sat Sun Mon Tue Wed Thu Fri SatWiFi-traffic ratio(a) WiFi- -traffic ratio (2013) heavylight 0 0.2 0.4 0.6 0.8 1Sat ... 0.8 1Sat Sun Mon Tue Wed Thu Fri SatWiFi-traffic ratio(b) WiFi- -traffic ratio (2015) heavylightFigure 7: WiFi- -traffic ratio (2013 and 2015).3.3.2 WiFi- -traffic ratio and WiFi- -user ratioWe introduce two metrics to highlight the benefit of ... ratioWe introduce two metrics to highlight the benefit of us-ing WiFi networks: The WiFi- -traffic ratio is defined as theWiFi download traffic volume divided ... traffic volume is downloaded via WiFi.The second metric is the WiFi- -user ratio, defined as the ra-tio of the number of ... as the ra-tio of the number of users associating to WiFi networks ineach time bin. A ratio close to 1.0 means ... at that time,most users are using WiFi.We see that the WiFi- -traffic ratio increases showing in-creasing use of WiFi o?oading. Figure 6(a) compares theWiFi-traffic ratio over each day of ... of the week in 2013 with2015. We observe that the WiFi- ... -traffic ratio varies between0.4 and 0.9 with clear diurnal trend. WiFi traffic is the high-est from 11pm to 2am and the ... lowest in weekday afternoons.This temporal pattern differs from the simple WiFi trafficvolume (Figure 2). Traffic penetration via WiFi is commonat night though cellular traffic is still dominant in ... data also confirms thatWiFi o?oad is increasing, with the mean WiFi- -traffic ratiogrowing to 0.71 in 2015 from 0.58 in 2013.WiFi- -user ratio shows that more users connect to WiFinetworks in ... toWiFi during peak hours. We examine the temporal variationin the WiFi- -user ratio in Figure 6(b). The number of userspeaks from ... to 2am, while 10am to 6pm is the off-peak.The mean WiFi- -user ratio also increases from 32% in 2013to 48% in ... hitters but also lightusers recognize the benefit of connecting to WiFi. . However,only 50% of devices use WiFi even in the peak time.3.3.3 Difference in usersWe now highlight ... time.3.3.3 Difference in usersWe now highlight the difference in the WiFi- -traffic ratioand WiFi- -user ratio between light users and heavy hitters.We emphasize that ... in 2015, heavy hitters o?oad most oftheir traffic volume to WiFi, , and that the ratio of o?oadtraffic for light users ... users also increases over the years. Figure 7represents changes in WiFi- -traffic ratio from 2013 to 2015.Heavy hitters already o?oad most ... 0.8 1Sat Sun Mon Tue Wed Thu Fri SatWiFi-user ratio(a) WiFi- -user ratio (2013) heavylight 0 0.2 0.4 0.6 0.8 1Sat ... 0.8 1Sat Sun Mon Tue Wed Thu Fri SatWiFi-user ratio(b) WiFi- -user ratio (2015) heavylightFigure 8: WiFi- -user ratio (2013 and 2015). 0 0.1 0.2 0.3 0.4 ... Sun Mon Tue Wed Thu Fri SatRatio of users(a) Android (2013)WiFi- -userWiFi-offWiFi-available 0 0.1 0.2 0.3 0.4 0.5 0.6Sat Sun Mon ... Sun Mon Tue Wed Thu Fri SatRatio of users(b) Android (2015)WiFi- -userWiFi-offWiFi-available 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8Sat ... Mon Tue Wed Thu Fri SatRatio of users(c) IOS (2013&2015) WiFi- -user (2013)WiFi- -user (2015)Figure 9: Ratio of users: (a) Android 2013, (b) ... typ-ical daily trend on weekdays.We see a large increase of WiFi- -user ratio especially forheavy-hitters during the three years. Figure 8 ... during the three years. Figure 8 illustratesthe increase of the WiFi- -user ratio. The difference betweenheavy hitters and light users is ... (mean: 51%). In 2015, morethan 80% of heavy hitters use WiFi during their peak time(mean: 68%). These results suggest that some ... with device OS. We first define more detailed cel-lular and WiFi usage to quantify the potential opportunityof WiFi o?oading. For Android devices, we categorize usersinto three groups, as ... Figure 9(a). The first categoryis users connecting their device to WiFi networks (WiFi- -user ; pink curve). The second category consists of WiFi- -offusers whose WiFi interfaces are off (blue curve). The thirdconsists of WiFi- -available users who use cellular interfacesas the primary link but ... cellular interfacesas the primary link but still turn on their WiFi interface(green curve). Note that the WiFi- -off users explicitly turn257off the WiFi
and the WiFi- -available users do notconnect to any WiFi ... AP.We confirm that 40% of Android devices explicitly turnoff their WiFi interface, but this ratio decreases during thethree years. The ratio ... this ratio decreases during thethree years. The ratio of the WiFi- -available users is sta-ble at around 0.25 in both 2013 ... thata quarter of the users could o?oad their traffic to WiFi ifany appropriate APs are discovered. We discuss these userslater in ... in the ratio ofWiFi-off users is opposite of that of WiFi users; the peaktime is business hours (10am-6pm). Our results show ... day in 2013 (blue curve). However, thesituation of the low WiFi usage improves in 2015, demon-strating that the ratio of WiFi- -off users drops from 50% to40% (Figure 9(b)).WiFi connectivity of iOS is higher than that of Android.For iOS ... devices, the software does not report detailed in-formation about the WiFi interface. However, we concludethat iOS devices connect to WiFi 30% more than do Androiddevices, as shown in Figure 9(c). ... 9(c). We also confirm that thereis no difference in the WiFi- -user ratios among three cellu-lar carriers providing iPhones. Thus, WiFi- -user ratio differsbetween the two device OSes rather than cellular ... the two device OSes rather than cellular carriers.3.4 Usage of WiFi networksNext we illustrate the results of user adoption of availableWiFi ... in offices, and in public places.3.4.1 Home, public, and other WiFi networksWe first define the locations of WiFi APs users connectedto. We identify each WiFi AP users associate with by its(BSSID, ESSID) pair (the MAC ... address of AP and its net-work name). We then categorize WiFi networks into thesethree types:Home: We identify home locations as the ... Japanese cellular providersfor their customers ( 1), or by free/commercial WiFi providers(e.g., 7Spot and Metro Free Wi-Fi). We categorize FONAPs that ... Other associated pairs are mainly located at of-fices or mobile WiFi APs. It also includes some open APsprovided by shops and ... between the two datasets.This data suggests that coverage of public WiFi networksis broad in the Greater Tokyo. The maps of public ... result suggests that users canpotentially connect to widely deployed public WiFi
... public, and office in2013 and 2015. The major contribution of WiFi traffic vol-ume is home networks (95% of the total volume). ... home networks (95% of the total volume). The publicand office WiFi traffic volumes are much smaller (4% of thetotal volume), though ... during this pe-riod. The diurnal patterns of public and office WiFi arein opposition to that of home WiFi. . These clear patternsin 2015 emphasize the increase of users ... volume [Mbps]Date (Mar. 2015, JST)Office TXOffice RX(f) office 2015Figure 11: WiFi traffic volume: (a) home in 2013, (b) public in 2013, ... (A), heavy hitters (H), and lightusers (L).3.4.2 Access patterns of WiFi networks: locationWe first show the number of associated APs increases ... light users (L). We find evidence that usersconnect to multiple WiFi APs in late years. In 2013, 70%of users only connect ... mobility patterns.Next we detail that the main usage patterns of WiFi net-works are still only at home, and that the use ... day for allusers. The HPO column indicates the possible combinationsof WiFi network types; home (H), public (P), and other (O)networks. For ... day. For users using oneAP, the percentage of using home WiFi decreases from 55%to 46%, though it is still dominant. About ... dominant. About 10% of the usersdo not have any home WiFi network but connect to otherWiFi networks. They have no broadband ... otherWiFi networks. They have no broadband wired links athome for WiFi o?oading; however, they are aware of WiFitraffic o?oading. For users ... further show that the distribution of user connectiondurations to one WiFi network do not change over time.Figure 13 shows the complementary ... any WiFinetwork type from 2013 to 2015.3.4.3 Access patterns of WiFi networks: 2.4 and 5GHzRecent WiFi APs operate in two frequency bands: 2.4 and5GHz. We find ... association time (hours)public office home2013-15 2013-15 2013-15Figure 13: CCDFs of WiFi connectionduration. 0 0.1 0.2 0.3 0.4 0.5 0.62013 2014 2015 ...
... -70 -60 -50 -40 -30 -20 -10PDFRSSIhomepublicFigure 15: PDFs of WiFi RSSI for asso-ciated APs (2015).ESSIDsHPO 2013 2014 2015per day1100 54.7% ... than20% at home and office APs are 5GHz, i.e., most WiFi APsmainly operate in 2.4GHz. For public WiFi, , we see thatmore than half are 5GHz APs in ... 5GHz in public location is due to recent aggressiverollout of WiFi, , while long device lifecycle means that homeand office have ... homeand office have not yet needed to upgrade.3.4.4 Quality of WiFi networks: signal strengthThere are several factors that affect the QoS/QoE ... signal, and an RSSI larger than ?70dBm isgenerally better for WiFi connectivity. For example, it hasbeen reported that the retransmission probability ... smaller RSSIs around ?60dBm.For narrowing the cell size in public WiFi, , the mean signalstrength can be weaker than that of ... the mean signalstrength can be weaker than that of home WiFi networks.As the result, we confirm that 12% of associated publicWiFi ... to deploy more 5GHz APs in publicWiFi networks.3.4.5 Quality of WiFi networks: channel usage 0 0.05 0.1 0.15 0.2 0.25 0.3 ... WiFinetworks is an important factor that affects the performanceof the WiFi networks. In an IEEE 802.11b/g/n network, 13channels are available in ... have a function fordetecting and avoiding channel interference. In addition,public WiFi providers design and deploy their APs to avoidcross channel interference ... devices. Devices connected mainly to Ch1, Ch6,and Ch11 in public WiFi APs, indicating that devices areset up to minimize interference. Such ... we observe improvement in avoiding possible in-terference in later years; WiFi channels at home are moredispersed and have less concentration on ... have less concentration on Ch1 in 2015.3.5 Availability of public WiFi
WiFi- -availableusers can o?oad to public WiFi networks, we next look atWiFi coverage, signal strength, and how ... theycan o?oad.We first suggest a wide coverage of potentially availablepublic WiFi APs. We sum up all detected public networks(all) and detected ... the downtownarea (Shinjuku, Shibuya area), for example, the number ofpublic WiFi networks with strong signal is over 10000 inone cell, while ... 5GHz APs in public networkshas largely improved.Next, we identify that WiFi- -available users ( 3.3.4) en-counter a few strong and available ... users ( 3.3.4) en-counter a few strong and available public WiFi APs, andthat the number of such APs increases. Figure 17 ... complementary cumulative distribution of the number ofdetected public APs per WiFi- -available device per 10 min-utes in 2015. The plot labeled ... fewer than 10 2.4GHz APs. As expected, the numberof strong WiFi networks is much smaller than all detectedWiFi networks. The data ... of strong2.4GHz and strong 5GHz APs resembles each other, sug-gesting WiFi APs operating in both bands.Finally, we infer that WiFi o?oading to available publicWiFi networks can reduce 15-20% of daily ... reduce 15-20% of daily cellular traffic forCell home Cell other WiFi home WiFi publictype % type % type % type %2013brows. 38.0 brows. ... users. As described in 1, cellular providersdeploy free public WiFi service to their customers. Deviceswhose WiFi interfaces are enabled but not associated withany WiFi networks (i.e., WiFi- -available user) can o?oadtheir cellular traffic to the public WiFi networks of theirproviders. Examining of such available WiFi networks forWiFi-available users, we confirm that 60% of WiFi- -availableusers have opportunities to connect to stable public WiFinetworks. We ... then concludethat 15-20% of daily cellular traffic volume for such WiFi- -available users can be transferred to public WiFi networks.3.6 Application breakdownWiFi o?oading often requires application support, so wenow ... an applica-tion).We distinguish application usage at home or other placeswith WiFi and cellular network interfaces. Table 6 and Ta-ble 7 show ...
... networks like WiFi.The ratio of video traffic increases especially in WiFi net-works over the years, while we see that most common ... (all)5GHz (strong)Figure 17: CCDFs of the number of de-tected public WiFi networks per deviceper 10 min. 0 0.2 0.4 0.6 0.8 ... 2015potentiallycappedothersFigure 19: Effect of soft bandwidth cap.Cell home Cell other WiFi home WiFi publictype % type % type % type %2013brows. 23.0 brows. ... Furthermore, we observed an in-crease in video traffic in public WiFi networks. Thus, usersdemand for public WiFi networks is changing from simpleconnectivity-sensitive to more bandwidth-consuming applica-tions. Instead, ... the productivity category increases in up-load volume. In particular, in WiFi networks at home, thiscategory accounts for a high percentage of ... dueto online storage software that uploads/downloads large filesonly if a WiFi interface is available. Thus, applications seemto play a major role ... available. Thus, applications seemto play a major role in promoting WiFi o?oading.Finally, for light users, the contribution of video applica-tions to ... not ranked in the top five categories in bothcellular and WiFi networks.3.7 Software updateWhile 3.6 looked at typical application use, ... by default. Thus, this event provides acase study of application-forced WiFi o?oading.The impact of the update on traffic volume is high; ... newversion would have a longer tail.We find that availability of WiFi is important to timely de-ploy updates; lack of WiFi can delay OS upgrades. In fact,only 14% of users without ... without homeWiFi seem to go out-of-their way to access public WiFi toobtain the update. We inspect 19 updated devices that haveno ... Eleven users without home AP up-date their devices via public WiFi, , and two via office APs.This observation is unexpected but ...
... of smart-phone traffic on residential broadband traffic, connectivityand quality of WiFi networks, and bandwidth-consumingapplications.4.1 Large impact of home WiFi offloadWe start by examining the impact of o?oaded traffic vol-ume ... the me-dian cellular download traffic volume is 36MB/day and themedian WiFi download traffic is 51MB/day in 2015 ( 3.2).Thus, 58% of ... 51MB/day in 2015 ( 3.2).Thus, 58% of smartphone traffic is WiFi; ; a 1.4:1 ratio ofWiFi-to-cellular traffic. This ratio is higher ... 28% (= 20%*1.4) of the total residential broadbandtraffic volume as WiFi traffic by smartphones because 95%of WiFi traffic is at home ( 3.4.1).We also compare smartphone traffic ... is 12% (= 51MB / 436MB).Thus, the growth in o?oaded WiFi traffic even for lightusers has had a large impact on ... hidden in the traffic volume of heavy hitters.4.2 Connectivity of WiFi networksTo understand the behavior of user that cannot be seenin ... 5.1 5.4 4.6 6.1 7.2 5.5Table 8: User survey: associated WiFi APs during the mea-surements (%: 2013/2014/2015).about WiFi APs to users. The first question is ?where didyou connect ... to users. The first question is ?where didyou connect to WiFi APs in three locations (home, office,and public)?? (Table 8), to ... 3.4.2). The second one is ?whydid you not connect to WiFi APs in these three locations??(Table 9), to understand possible causes ... for office network is low and stable. The percentages ofhome WiFi APs are consistent with our estimation (Table 5),but, high percentages ... consistent with our estimation (Table 5),but, high percentages of public WiFi differ from our esti-mation. This gap represents the difference between ... think they havemore connectivity than they really do in public WiFi
However, the results of the associated WiFi networksdemonstrate that the percentage of users with multiple APshas been ... 9), we summarize the plausible reasonsfor the low availability of WiFi networks during daytimeas follows. (1) Users report that connecting private ... daytimeas follows. (1) Users report that connecting private smart-phones to WiFi networks at offices (Bring your own device;BYOD) is still not ... Table 9. In addition, a major reason of ?Other? foroffice WiFi networks is due to the security policy that doesnot allow ... doesnot allow employees to connect their personal smartphoneto the office WiFi ... networks. Our traffic data showed a sig-nificant increase in office WiFi traffic volume (Figure 11)though the number of inferred office APs ... office APs (Table 4) remainsstable during the three years.(2) Configuring WiFi APs is difficult. Configuration prob-lems decrease in public WiFi networks over time in the sur-vey. Users report in our ... their mobile devices is often complicated.To reduce complexity, since 2013 WiFi APs from cellularproviders use SIM-based authentication [25] without userand password ... userand password information. As a result, smartphone cus-tomers can use WiFi APs from their provider with no man-ual actions. Coupled with ... believe this simplification in their usehelps increase the degree of WiFi o?oad.(3) Many users have no incentive to use public WiFi net-works because the current quality of cellular networks isenough and ... be o?oaded if users ap-propriately configure the settings for public WiFi networks( 3.5).(4) Finally, we found that security with public WiFi wasa significant concern, and that battery life was not.263Reason home ... 9 5 4Table 9: User survey: reasons for unavailability of WiFi APs(%). Multiple answers were allowed.4.3 Quality of public WiFi networksWe discuss how public WiFi networks can improve theirquality. The quality and accessibility of WiFi ... networks havebeen improved over the years. However, 12% of associatedpublic WiFi networks using 2.4GHz are characterized by alow RSSI ( 3.4.4). ... problem, deviceOSes should implement tests for the quality of a WiFi net-work before starting to o?oad to prevent the degradation ofuser ... Tomitigate performance degradation due to channel interfer-ence among different public WiFi networks, deployment ofAPs that support multiple providers by announcing multi-ple ... will be also cost-effective to use theseAPs for providing free WiFi APs to foreign visitors insteadof deploying new APs, towards the ... APs has already been discussed inthe context of providing free WiFi in disaster areas after theTohoku earthquake.We also observed a significant ... more robust againstnoise and helpful in improving the quality of WiFi network[40, 46]. This migration also contributes to increases in avail-ability, ... contributes to increases in avail-ability, use, and traffic in public WiFi networks. Further-more, the limited spatial and more reliable nature of ... large growth of bandwidth-consuming ap-plication traffic ( 3.6). Users with WiFi access downloadmore video (Table 7), suggesting that users understand thatrich-bandwidth ... video (Table 7), suggesting that users understand thatrich-bandwidth and low-cost WiFi networks are available,and changing their usage accordingly. Similarly, other bandwidth-consuming ... changing their usage accordingly. Similarly, other bandwidth-consuming applications contribute to WiFi o?oading. Forexample, iOS shows a pop-up message to users to ... shows a pop-up message to users to promotethe use of WiFi for software updates. We see a similarchange in upload use ... software updates. We see a similarchange in upload use when WiFi is available. For example,WiFi has been largely used for the productivity category,which includes online ...
... another adap-tive behavior of users. Some users rely on public WiFi net-works for their software updates ( 3.7). Furthermore, weobserved an ... demand is towards casual use of bandwidth-consumption applications in public WiFi networks. Thistrend may introduce a change of the economic cost ... WiFinetworks, because low-cost connectivity is the main advan-tage with current WiFi networks.5. RELATED WORKCellular smartphone usage: Quantifying the diverseusage of smartphones ... that time-dependent pricing helps reduce peaktraffic volume in cellular networks [30].WiFi smartphone usage: Some studies focused on WiFismartphone usage in campus ... in public transportations [24]. Theyargued that typical applications of such WiFi networks wereHTTP, more specifically video, social networking, and soft-ware updates. ... social networking, and soft-ware updates. In addition, the deployment of WiFi APs andtheir interference are discussed [2].Throughput/delay performance of WiFi and 3G/LTEnetworks: Several performance metrics have been com-pared using vehicle-based ... [45]measurements. They showed that the throughput and de-lay performance of WiFi networks are basically better thanthose of cellular networks, and they ... evaluated the performance gain ofWiFi o?oading on the basis of WiFi APs discovered usingvehicle based measurements [31, 23].WiFi networks: Home WiFi usage has been analyzedin detail [33, 39, 13, 14]. An ... traffic ra-tio [33], and the availability [13, 14] of home WiFi networkshave been intensively studied. A large analysis of ESSIDsfocused on ... the social relationships of users [5]. A recentlarge-scale survey of WiFi networks revealed application us-age and channel interference in the wild ... to pro-vide a comprehensive view of smartphone usage in cellular264and WiFi networks from measurements with over 1500 users,for better understanding the ... 1500 users,for better understanding the adoption of smartphone usersto avaialble WiFi networks. We previously explained thepreliminary results of WiFi traffic o?oading performancebased on a two-day measurement with 450 Android ... dif-ferences regarding device OS, location, and applications.6. CONCLUDING REMARKSWe measured WiFi and 3G/LTE smartphone usage frommore than 1500 Android/iOS smartphone users ... 2013, 2014, and 2015. Ourresults showed that smartphone users select WiFi and cel-lular networks based on several tradeoffs. In particular, wehighlighted ... tradeoffs. In particular, wehighlighted slow but clear adoption of available WiFi envi-ronment at home, in offices, and in public spaces. Lightusers ... in public spaces. Lightusers have been o?oading their traffic to WiFi networks,while heavy hitters totally rely on WiFi networks for theirtraffic o?oad. Moreover, we confirmed that wide deploy-ment ... for theirtraffic o?oad. Moreover, we confirmed that wide deploy-ment of WiFi APs enables users to obtain more opportu-nities to o?oad their ... users to obtain more opportu-nities to o?oad their traffic to WiFi networks. Users whohave access to public WiFi send more traffic, and WiFi-
... required forlarge software updates. This adoption indicates that usersrecognize public WiFi networks as a part of the commonnetwork infrastructure.AcknowledgmentsWe thank our ... M. Voelker, and A. C.Snoeren. Analysis of a mixed-use urban WiFi network:When metropolitan becomes neapolitan. In IMC?08,pages 85?98, Vouliagmeni, Greece, Oct ... W. Wei.Network performance of smart mobile handhelds in auniversity campus WiFi network. In IMC?12, pages315?328, Boston, MA, 2013.[9] K. Cho. Broadband ... Hou, and S. R. Das. Performancecomparison of 3G and metro-scale WiFi for vehicularnetwork access. In IMC?10, pages 301?307, Melbourne,Australia, Nov 2010.[13] ... S. Banerjee. Beyonddeployments and testbeds: Experiences with publicusage on vehicular WiFi hotspots. In MobiSys?12,pages 393?405, Low Wood Bay, UK, Jun 2012.[25] ...
... S. Chong, and Y. Yi. Mobiledata o?oading: How much can WiFi deliver? InCoNEXT?10, page 12, Philadelphia, PA, Dec 2010.[32] X. Liu, ... Patro, S. Govindan, and S. Banerjee. Observinghome wireless experience through wifi aps. InMobiCom?13, pages 339?350, Miami, FL, Sep 2013.[41] M. Rodrig, ... CA, Jun 2011.[45] J. Sommers and P. Barford. Cell vs. WiFi: : On theperformance of metro area mobile connections. InIMC?12, pages ... traffic behaviorDaily user traffic volumeWiFi vs. CellularAggregated viewWiFi-traffic ratio and WiFi- -user ratioDifference in usersDifference in device OSUsage of WiFi networksHome, public, and other WiFi networksAccess patterns of WiFi networks: locationAccess patterns of WiFi networks: 2.4 and 5GHzQuality of WiFi networks: signal strengthQuality of WiFi networks: channel usageAvailability of public WiFi ... APsApplication breakdownSoftware updateEffect of soft bandwidth capImplicationsLarge impact of home WiFi offloadConnectivity of WiFi
Quality of public WiFi

Title:
Characterizing WiFi connection and its impact on mobile users: practical insights

Keywords:
WiFi
WiFi performance

Abstract:
<p>Smartphones and WiFi networks are becoming pervasive. As a result, new applications and ... applications and services are being offered to smartphone users through WiFi networks. Some of the more novel applications provide data services ... applications provide data services to pedestrians as they move through WiFi coverage areas in public loca- tions such as railway stations. ... the connection set-up time. In this paper we characterize the WiFi connection set-up process. Using data from voluntary Android smartphone users, ... Using data from voluntary Android smartphone users, we show that WiFi connection setup have significant delays, sometimes as high as 10s. ... this is due to losses of DHCP messages at the WiFi access point. We also show that some of the methods ... manufactures are suboptimal and this can be addressed at the WiFi access point. Finally using this insight we extend a known ... known mathematical model, which will help in the dimensioning of WiFi

References:
J. Eriksson, H. Balakrishnan, and S. Madden. Cabernet: Vehicular content delivery using wifi. In Proceedings of the Annual International Conference on Mobile Computing and Networking, MOBICOM, pages 199--210, 2008.
S. Liu and A. Striegel. Casting doubts on the viability of wifi offloading. In Proceedings of the 2012 ACM SIGCOMM workshop on Cellular networks: operations, challenges, and future design, pages 25--30. ACM, 2012.
R. Schwartz and M. Johansson. Carrier WiFi Offload. http://www.wireless2020.com/docs/CarrierWiFiOffloadWhitePaper03202012.pdf, 2012.

Full Text:
Characterizing WiFi connection and its impact on mobile users: practical insightsCharacterizing WiFi Connection and Its Impact on MobileUsers: Practical InsightsSuranga SeneviratneSchool of ... MohapatraDepartment of ComputerScience, University ofCalifornia, DavisCA, 95616prasant@cs.ucdavis.eduPierre-Ugo TournouxNICTA, Australiatournoux@gmail.comABSTRACTSmartphones and WiFi networks are becoming pervasive.As a result, new applications and services ... new applications and services are being offeredto smartphone users through WiFi networks. Some of themore novel applications provide data services to ... novel applications provide data services to pedestriansas they move through WiFi coverage areas in public loca-tions such as railway stations. One ... is the connection set-up time. In this paper wecharacterize the WiFi connection set-up process. Using datafrom voluntary Android smartphone users, we ... device manufactures are suboptimaland this can be addressed at the WiFi access point. Finallyusing this insight we extend a known mathematical ... a known mathematical model,which will help in the dimensioning of WiFi networks forpedestrian smartphone users.Categories and Subject DescriptorsC.4 [Computer-Communication Networks]: NetworkArchitecture ... users.Categories and Subject DescriptorsC.4 [Computer-Communication Networks]: NetworkArchitecture and Design?Wireless communicationKeywordsWiFi; WiFi Performance; Smartphones1. INTRODUCTIONThe proliferation of mobile devices is enabling a ... be downloaded byusers when they move through areas covered by WiFi net-works at public locations such as railway stations, shoppingmalls, and ... mobile device tobe able to establish a connection with a WiFi access point,and download a minimum quantity of data during the ... dimensioned [13].The former will depend on the connection establishmentprocess of WiFi networks. Although this has been stud-ied extensively from a handoff ... recently been reported that operational characteristicsof smartphones when connected to WiFi networks differ sig-nificantly from other systems [8, 11]. This paper ... used for data dissemination through shortlived connections, by characterizing the WiFi connection set-up time and its impact when using commercial off-the-shelfsmartphones. ...
... connection set-up delay is due to the messagelosses at the WiFi access points. Because of this, weshow that previously proposed solutions ... as follows. Sec-tion 2 details the various phases of the WiFi connection pro-cess while Section 3 describes our experimental setup andthe ... Section 8, we conclude the study.2. BACKGROUNDData transfer in a WiFi network involves a connection set-up and the transfer of data. ... probe request for each AP it has connectedpreviously on each WiFi channel. The frequency of the ac-tive probe request depends on ... The frequency of the ac-tive probe request depends on the WiFi card of the mobiledevice. When active probing is used, the ... the signalstrength is insufficient to associate with the AP.? The WiFi subsystem of a smartphone sleeps and wakesup periodically to save ... for the following two reasons. Firstly, as the APsin public WiFi networks will transmit a beacon at regularintervals and the users ... and the data transfer can commence.3. EVALUATION METHODOLOGYAnalyzing the smartphones? WiFi connection establish-ment requires distinguishing between the three phases ofconnection set-up, ...
... Field DatasetField Dataset was obtained by installing an applicationwe developed, Wifi Event Monitor1 on volunteers? smart-phones. Wifi Event Monitor runs in the background andlogs the WiFi connectivity events explained in Section 2,with timestamps during devices? everyday ... the Airpcap Nx3. A number of commod-ity Android phones with Wifi Event Monitor were used toperiodically connect to a number of ...
... to previous work that im-plied the long DHCP times in WiFi networks were due tothe loss of DHCP Discover and DHCP ...
... use the following distance throughput model obtainedfrom [4] for 802.11n WiFi networks.S(rn) ={(ar2 + br + c) if 0 < r ... the networks are servicing a small user population.However in public WiFi networks where the networks haveto serve large user populations, most ...
... By-chkovsky et al. [2] measured the connection set-up timesto available WiFi access points while driving around Bostonand Seattle and reported that ... believe , especially thelease validity will be low in public WiFi networks.Eriksson et al. [5] and Soroush et al. [14] proposed ... wireless APs. As described in Section 2.1 we expectthe public WiFi networks to transmit a beacon at regularintervals and the smartphones ... are found to be in range.It might be argued that WiFi proposals such as WiFi Di-rect [1] might address this issue. However, WiFi Direct isstill in its early ages of development and its ... points, we showed that one ofthe three components of the WiFi connection set-up pro-cess, namely the IP address acquisition time, on ... ad-dress acquisition delay is the discard of DHCP messages atthe WiFi access point than these messages being lost on thewireless link. ... that these solutions arenot optimal for the use in public WiFi hotspots to distributecontent to pedestrians. Finally, through experimentation,we showed that ... calculate the downloadable data volumefor pedestrian users while walking across WiFi coverage ar-eas. The analysis of the connection set-up time and ... time and themathematical model provides useful insights for dimension-ing public WiFi networks for pedestrian users.9. REFERENCES[1] Wi-Fi Peer-to-Peer (P2P) Specification v1.1.http://www.wi-fi.org/Wi-Fi_Direct.php, ... Eriksson, H. Balakrishnan, and S. Madden.Cabernet: Vehicular content delivery using wifi. . InProceedings of the Annual International Conference onMobile Computing and ...
... S. Liu and A. Striegel. Casting doubts on the viabilityof wifi o?oading. In Proceedings of the 2012 ACMSIGCOMM workshop on Cellular ... area, volume 6640 LNCS.2011.[13] R. Schwartz and M. Johansson. Carrier WiFi O?oad.http://www.wireless2020.com/docs/CarrierWiFiOffloadWhitePaper03202012.pdf, 2012.[14] H. Soroush, P. Gilbert, N. Banerjee, M. D. ...

Title:
Cell vs. WiFi: on the performance of metro area mobile connections

Keywords:
WiFi

Abstract:
<p>Cellular and 802.11 WiFi are compelling options for mobile Internet connectivity. The goal of ... conditions. In this paper, we compare and contrast cellular and WiFi performance using crowd-sourced data from Speedtest.net. Our study considers spatio-temporal ... 15 week period. Our basic performance comparisons show that <i>(i)</i> WiFi provides better absolute download/upload throughput, and a higher degree of ... throughput, and a higher degree of consistency in performance; <i>(ii)</i> WiFi networks generally deliver lower absolute latency, but the <i>consistency</i> in ... depending on the particular access type <i>e.g.</i>, HSPA, EVDO, LTE, WiFi, , <i>etc</i>.) and service provider. More broadly, our results show ... broadly, our results show that performance consistency for cellular and WiFi is much lower than has been reported for <i>wired</i> broadband. ... broadband. Temporal analysis shows that average performance for cell and WiFi varies with time of day, with the best performance for ... are subregions that offer consistently better performance for cell or WiFi. . Comparisons between metro areas show that larger areas provide ...

References:
A. Balasubramanian, R. Majahan, and A. Venkataramani. Augmenting Mobile 3G Using WiFi. In Proceedings of ACM MobiSys '10, San Francisco, CA, June 2010.
V. Birk, S. Rayanchu, S. Saha, S. sen, V. Shrivastava, and S. Banerjee. A Measurement Study of a Commercial-grade Urban WiFi Mesh. In Proceedings of ACM Internet Measurement Conference, Vouliagmeni, Greece, October 2008.
P. Deshpande, X. Hou, and S. Das. Performance Comparison of 3G and Metro-Scale WiFi for Vehicular Network Access. In Proceedings of ACM Internet Measurement Conference, Melbourne, Australia, November 2010.

Full Text:
imc099-sommers.dviCell vs. WiFi: : On the Performance of Metro Area MobileConnectionsJoel SommersColgate Universityjsommers@colgate.eduPaul ... Metro Area MobileConnectionsJoel SommersColgate Universityjsommers@colgate.eduPaul BarfordUniversity of Wisconsinpb@cs.wisc.eduABSTRACTCellular and 802.11 WiFi are compelling options for mobile Inter-net connectivity. The goal of ... use conditions. In this paper, we compare andcontrast cellular and WiFi performance using crowd-sourced datafrom Speedtest.net. Our study considers spatio-temporal per-formance ... 15 week period. Our basic per-formance comparisons show that (i) WiFi provides better absolutedownload/upload throughput, and a higher degree of consistency ... absolutedownload/upload throughput, and a higher degree of consistency inperformance; (ii) WiFi networks generally deliver lower absolutelatency, but the consistency in latency ... widely depending on theparticular access type (e.g., HSPA, EVDO, LTE, WiFi, , etc.) andservice provider. More broadly, our results show that ... More broadly, our results show that performanceconsistency for cellular and WiFi is much lower than has been re-ported for wired broadband. ... wired broadband. Temporal analysis shows that averageperformance for cell and WiFi varies with time of day, with the bestperformance for large ... are subregions that offer consistently better per-formance for cell or WiFi. . Comparisons between metro areas showthat larger areas provide higher ... and growingdemand for bandwidth by mobile users.A vexing problem for WiFi enabled cell phone users, serviceproviders and application designers is seeking ...
... study is to understand the spatio-temporal char-acteristics of performance of WiFi- -enabled cell phones in a selec-tion of metro areas with ... questionssuch as: what is the relative performance of cellular vs. WiFi ina given geographic area? How does performance vary across localaccess ... performance vary across localaccess providers, and how does cell and WiFi performance vary insub-regions within the metro area? How does cellular ... work is to for-mulate conclusions about the spatio-temporal aspects of WiFi en-abled cell phone performance that will lead to improvements in ... mobile devices in a metro area.As expected, we find absolute WiFi download and upload perfor-mance to be superior to cellular performance ... to be superior to cellular performance in most areas, andthat WiFi ... exhibits a higher degree of performance consistency. Wealso find that WiFi latency measurements are at least a factor of twolower than ... cellular access. Further, the absolute la-tency difference between cell and WiFi tends to be smaller in largermetro areas and the overall ... access technologies such as 4G LTE offer throughputspeeds comparable to WiFi, , the upload performance consistency iscurrently low, suggesting that these ... generally, our results show that performance con-sistency for cellular and WiFi is significantly lower than has beenreported for wired broadband access. ... cycle but is highly variable. Specif-ically, we find that while WiFi performance tends to be more uni-form across subareas, cell performance ... be more localized in large metro areas forboth cell and WiFi. . These results have implications for both usersand operators in ...
... (market type) Pop. Metro Rank Annual PCI iOS AndroidUnique # WiFi # cell Unique # WiFi # cellhandsets tests tests handsets tests testsNew York, NY (large) ... download speeds (in kb/s), latency (in milliseconds), accesstype (cellular or WiFi) ), and the cellular carrier or WiFi networkprovider. In the Android data set, for some tests we ... exists; we only know whether the access is viacell or WiFi. . For each of the apps, we also have a ... corpus, yet one that provides a broad perspective on cellularvs. WiFi
... initiated per day from cellular access,and 15,521 per day using WiFi. . Interestingly, for the Android app,there are 11,273 tests per ... tests per day on average via cellular technology,and 4,380 via WiFi, , while for the iOS app, there are only 2,464 ... testsvia cellular technology per day on average, compared with 11,141via WiFi. . Also, in 9,230 cases, we observe the same handset ... Lawrence,KS servers during the 15 week test period. Maps of WiFi clientlocations from these metro regions, and maps of client locationsfrom ... areas have similar profiles. In the large metroareas, cellular and WiFi tests are conducted with more uniformityover the highly populated metro ...
... cases, there is a high degree of overlap betweencellular and WiFi test locations.2.3 DiscussionWe argue that the Speedtest data are compelling ... our data via unique device identifiers, our comparisons betweencellular and WiFi performance are at an aggregate level. In futurework, we intend ... level. In futurework, we intend to carefully examine cellular versus WiFi perfor-mance on an individual user basis.We are limited, to a ... not have up-to-date ground truth locations of all celltowers and WiFi access points that provided connectivity for hand-sets for all tests. ... future evalua-tions of our data. There are similar archives for WiFi access pointse.g., [8], but the completeness of these databases is ... approach to assessing the spatio-temporal performance characteristics of cellular and WiFi through-put and latency in the target metro areas. This section ...
... metro region, we separate the series of testsby access technology (WiFi, , cell, or some more detailed cell ac-cess type) and ... one of its most well-known empirical characteristics. Prior studies of WiFi networks(e.g., [22]) and cellular traffic (e.g., [20]) have shown that ... diurnal pattern andhow the expected diurnal use of cellular and WiFi has an impact onperformance in our target metro areas. By ... typically comprise hundreds of square miles,potentially thousands of cellular and WiFi access points and mil-lions of users. As indicated above, this ...
... report the results of our spatio-temporal anal-yses of cellular and WiFi performance in the 15 target metro areas.305Table 3: Download throughput ... 15 target metro areas.305Table 3: Download throughput for cell and WiFi from the 15 target metro areas for full 15 week ... week period. All values are in kb/s.Location Cell Mean (Stdev) WiFi Mean (Stdev) Cell 5th% Cell Median Cell 95th% WiFi 5th% WiFi Median WiFi 95th%New York, NY 3194.4 (4234.7) 7621.7 (5574.8) 108.0 1678.0 12922.0 ... 2595.0 90.0 975.0 10789.0Table 4: Upload throughput for cell and WiFi from the 15 target metro areas for full 15 week ... week period. All values are in kb/s.Location Cell Mean (Stdev) WiFi Mean (Stdev) Cell 5th% Cell Median Cell 95th% WiFi 5th% WiFi Median WiFi 95th%New York, NY 1804.6 (4577.9) 2873.2 (3314.6) 52.0 772.0 5428.0 ... PerformanceOur analysis begins by examining the general characteristics ofcellular and WiFi performance in each of the target metro areas.These characteristics can ... Tables 3, 4, and 5. The side-by-side comparison shows that WiFi provides better maximum andaverage performance for nearly all regions for ... thatcause the average and 95th percentile to be higher than WiFi. . Thesetests are all from devices using the 4G LTE ... The tables also show that the difference in uploadperformance between WiFi and cell is much smaller than the dif-ference in download ... scatterplots of upload versus download per-formance for cell (left) and Wifi (right) for the Madison, WI metroarea. Each data point is ...
metro ar-eas. First, as with Tables 3 and 4, WiFi performance is generallyhigher than cell. We note that the highest ... arefor the LTE access technology. We also observe that for WiFi ac-cess, there are more obvious ?tiered? performance bands evident,especially for ... of these evident performance tiers in the upload di-rection. For WiFi networks, these bands likely represent differentservice plans available to customers. ... bands are most evident in the upload direction; es-pecially for WiFi, , there are no obvious download throughput tiers.We hypothesize that ... computed for each localaccess carrier. The plots are created from WiFi measurements only;306Table 5: Latency for cell and WiFi from the 15 target metro areas for full 15 week ... week period. All values are in milliseconds.Location Cell Mean (Stdev) WiFi Mean (Stdev) Cell 5th% Cell Median Cell 95th% WiFi 5th% WiFi Median WiFi 95th%New York, NY 282.0 (575.9) 111.9 (261.8) 68.0 159.0 786.0 ... Scatterplots of upload versus download performance for cellular (left) and WiFi (right) for the Madison, Wisconsin metroarea. Data points represent 95th ...
WiFi performance. Celllatencies are generally longer than WiFi, , with mean cell latenciesapproaching or exceeding a third of ... me-dian cell latency is at least twice as large as WiFi latency for nearlyall regions we consider (Columbia, SC is the ... we plot in Figure 5 empir-ical cumulative distribution functions for WiFi connections and forspecific cellular access technologies, for providers from which ... Scatterplots of iOS versus Android download and latency performance for WiFi. . The left plot shows median downloadperformance for iOS devices ...
... users on wired networks. In Figure 6, wealso observe that WiFi upload/download performance is generallymore consistent than cell upload/download, though it ... someproviders use the same network ?backhaul? infrastructure for bothcellular and WiFi access in an effort to optimize their network in-frastructures to ... cause ofthe observed similarity in performance consistency between cellu-lar and WiFi. . We intend to examine this hypothesis in detail infuture ... initially rolledout.With respect to latency performance consistency, we see thatwhile WiFi offers generally higher absolute throughputs and moreconsistent throughput, cellular latency ... poor performance consistency for most ac-cess types it offers (AS21928), WiFi latencies exhibit a lower de-gree of consistency than cellular access ... right), 2 of the 4 least consistent access tech-nologies are WiFi. . We hypothesize that this lower degree of per-formance consistency ... overprovisioning of buffers [21, 33], andit is likely that user WiFi access is often through a home-graderouter connected to a wired ... lower degree of performance consistency in WiFiis higher contention for WiFi frequency bands, and differences be-tween WiFi and cellular medium access control. However, sincewe observe the same ... However, sincewe observe the same pattern of lower consistency in WiFi acrossall metro areas?even the most sparsely populated ones where wewould ...
... exhibitsimilar qualities due to using the same backhaul infrastructure.Main findings.Absolute WiFi performance is better than cellular access, in gen-eral. Throughput performance ... buffer-ing or APIs that are suboptimally designed. Performance consis-tency of WiFi throughput is generally better than cellular, but cel-lular latency performance ... performance tends to be more consistent than WiFi.Lower consistency of WiFi latency is likely due to the impact ofoverbuffering at broadband ... marketsthere are similar numbers of invocations of the app over WiFi andcellular access (e.g., New York, NY and Los Angeles, CA), ... temporal characteristics of upload and down-load performance for cellular and WiFi. . Figure 8 shows the hourlyaverage upload and download performance ... areatypes (notice the different y-axis scales for cellular (top) and WiFi( (bottom)). We observe that for cellular access, the performancefor all ... has a clear impact on throughput performance. We observethat for WiFi connections, while the smallest metro areas have gen-erally lower throughputs, ... the differences are not as great among themetro areas for WiFi as they are for cellular connections.In other results (not shown ...
... performance for cellular downloads (top left), cellular uploads (top right), WiFi downloads (bottom left)and WiFi uploads (bottom right) for exemplars in each of the metro ... visually comparable.Figure 9 shows plots of interpolated upload performance for WiFi( (top) and cell (bottom) for three different metro regions (Chicago,IL, ... are subareas with consistently poor perfor-mance.We also observe that although WiFi upload performance is broadly,on average, at least twice as fast ... examining the Manchester, UK plots, we observe a clearseparation in WiFi upload performance between areas closer tothe city center and surrounding, ... examine spatial performance character-istics, and utilize available cell tower and WiFi maps, along witheconomic and population data, to drill down on ... center. Observed performancedifferences are likely due to cellular tower and WiFi base stationplacement, and density of placements, as well as local ... contentiondue to load.311Figure 9: Inverse distance weighting interpolation plots for WiFi upload performance (top) and cellular (bottom) for Chicago, IL(left), Manchester, ... body of work that examines the be-havior and characteristics of WiFi networks. Studies that are mostclosely related to ours have been ... et al. analyze the behaviorand characteristics of a city-wide commercial WiFi mesh networkin [13]. Their study is based on a diverse ... measurement of wide-area wireless networks [30]. Similar empirical studies of WiFi be-havior in localized settings include [9, 15, 22, 27]. More ... and Balakrishnan report results of an empirical study of110 different WiFi mesh networks in diverse markets around theworld [25]. While their ... all of the aforementioned studies expand the body ofknowledge on WiFi
... WiFiperformance simultaneously. The notion of the combined use ofcellular and WiFi in a vehicular setting is addressed in [10]. Thatstudy considers ... vehicular setting is addressed in [10]. Thatstudy considers cellular and WiFi availability in three cities as thebasis for their work. In ... for their work. In [17], Deshpande et al. evaluate cellularand WiFi performance in the New York metro area, also in a ... of cellular versus the lower availabil-ity, higher performance characteristics of WiFi. . While there aresimilarities between these studies and our own, ... using a larger body of crowd-sourced data.31205/010100200300400500600Testsperhourcellwifi05/010100200300400500600700TestsperhourcellwifiFigure 7: Cellular and WiFi tests conducted per hour in NewYork, NY (top) and Manchester, ... 2011 to 8 May, 2011.6. SUMMARY AND CONCLUSIONSCellular and 802.11 WiFi are the de facto connectivity optionsfor today?s mobile users. The ... performance. The raw comparison between the two tech-nologies shows that WiFi provides superior download performance,with maximum WiFi performance varying widely. The differencein upload performance is much smaller, ... much smaller, yet is also highly variable.We also find that WiFi latency measurements are at least a factorof two lower than ... thelargest metro areas, with performance decreasing for both cellu-lar and WiFi during the hours of peak use. Comparisons betweenmetro areas shows ... of our study is on a broad comparison ofcellular and WiFi in metro areas, our current results suggest severalconclusions about mobile ... our current results suggest severalconclusions about mobile performance. First, while WiFI offerssuperior download performance, the relatively predictable level ofcellular performance for ... performance would be en-hanced by further build out of both WiFi
... related datasets such as weatherconditions during test periods and cell tower/WiFi access pointlocations. We also plan to investigate hypotheses posed in ... A. Balasubramanian, R. Majahan, and A. Venkataramani.Augmenting Mobile 3G Using WiFi. . In Proceedings of ACMMobiSys ?10, San Francisco, CA, June ... V. Shrivastava, andS. Banerjee. A Measurement Study of a Commercial-gradeUrban WiFi Mesh. In Proceedings of ACM InternetMeasurement Conference, Vouliagmeni, Greece, October2008.[14] ... X. Hou, and S. Das. Performance Comparisonof 3G and Metro-Scale WiFi for Vehicular Network Access.In Proceedings of ACM Internet Measurement Conference,Melbourne, ...
Summary Review Documentation for?Cell vs. WiFi: : On the Performance of Metro Area MobileConnections?Authors: J. Sommers ... bandwidth testing site, to compare the per-formance of Cellular and WiFi networks. The data set used in thispaper is great, and ... what was presented. For example, it is wellknown that (1) WiFi has better upload/download rates than 3G, but3G latency is more ... upload/download rates than 3G, but3G latency is more stable than WiFi, , (2) WiFi/ /3G throughput islower than wired broadband and (3) 3G performance ... is poorer dur-ing peak hours, and (4) diurnal characteristics of WiFi/ /3G access.These are the major findings of this paper, but ... 15-weeks ofSpeedtest.net crowdsourced data of network performance for both3/4G and WiFi networks in several major metros. They compareperformance across providers and ... they can be used to useful effect. The mosttempting aspect?comparing WiFi to 3G?is not possible with thisdataset due to the way ... a high level, the title of your paper ?Cell vs. WiFi? ? is mislead-ing. It is extremely temping to try to ... you did refrain from conducting any quantitative analysis ofCell vs. Wifi, , your text and especially intro and conclusions arelittered with ... of the major metros studied) to try and determinewhether the WiFi and Cell performance at a given location for agiven device ... of Speedtest data withthe aim to understand performance differences between WiFi andcellular for 15 different geographical regions. The authors are ableto ...
... The authors are able to study the performance ofwireless access (WiFi and cellular) across time, space, and device.It would actually be ... one should expect.Actually, this kind of high upload speeds over WiFi (Fig. 3(b))make me wonder if you are actually measuring enterprise ... from from speedtest.net, acrowd-sourced performance measurement tool. The paper focuseson cellular/wifi performance measurements to cast light on the rel-ative performances these ... these complementary technologies. Some ofthe paper?s findings are obvious (i.e. WiFi offers lower RTTs andhigher throughput than cellular networks), but some ... to authors: You mention that the reason for less con-sistent wifi latencies is over-buffering at access points. That soundslike a secondary ... best - another explanation is the muchhigher contention in the WiFi frequency bands coupled with the?random? nature of wifi access control (backoffs, etc). In contrast,cellular networks make allocation decisions ...
... #5Summary: The paper uses crowd-sourced measurement data tocontrast cellular and WiFi performance along multiple dimensions.The paper does a good job of ... have been used in other studies that exam-ine cellular and WiFi performance. We argue that the fact that ourwork replicates some ... analysis of the possibility of usingLTE as a replacement for WiFi. . These are excellent suggestions,which we had already been considering. ...

Title:
WiFi-Nano: reclaiming WiFi efficiency through 800 ns slots

Keywords:
WiFi

Abstract:
<p>The increase in WiFi physical layer transmission speeds from 1~Mbps to 1 Gbps has ... 1500 byte packet from 12 ms to 12 us. However, WiFi MAC overheads such as channel access and acks have not ... 150 us on average per packet. Thus, the efficiency of WiFi has deteriorated from over 80% at 1 Mbps to under ... under 10% at 1 Gbps.</p> <p>In this paper, we propose WiFi- -Nano, a system that uses 800 ns slots} to significantly ... a system that uses 800 ns slots} to significantly improve WiFi efficiency. Reducing slot time from 9 us to 800 ns ... reception, thereby reducing ack overhead. We validate the effectiveness of WiFi- -Nano through implementation on an FPGA-based software defined radio platform, ...

Full Text:
Main.dviWiFi-Nano: Reclaiming WiFi Efficiency Through 800 nsSlotsEugenio Magistretti ?Rice UniversityTexas, USAemagistretti@rice.eduKrishna KantChintalapudiMicrosoft Research ... Indiakrchinta@microsoft.comBozidar RadunovicMicrosoft ResearchCambridgeCambridge, U.Kbozidar@microsoft.comRamachandran RamjeeMicrosoft Research IndiaBangalore, Indiaramjee@microsoft.comABSTRACTThe increase in WiFi physical layer transmission speeds from 1 Mbpsto 1 Gbps has ... a 1500 byte packetfrom 12 ms to 12 s. However, WiFi MAC overheads such aschannel access and acks have not seen ... about 150 s on average per packet. Thus,the efficiency of WiFi has deteriorated from over 80% at 1 Mbps tounder 10% ... Mbps tounder 10% at 1 Gbps.In this paper, we propose WiFi- ... -Nano, a system that uses 800 nsslots to significantly improve WiFi efficiency. Reducing slot timefrom 9 s to 800 ns makes ... packet reception, thereby reducing ackoverhead. We validate the effectiveness of WiFi- -Nano through im-plementation on an FPGA-based software defined radio platform,and ... Las Vegas, Nevada, USA.Copyright 2011 ACM 978-1-4503-0492-4/11/09 ...$10.00.General TermsDesign,Experimentation1. INTRODUCTIONWhile WiFi physical layer (PHY) data rates have increased from1 Mbps in ... reducing the channel ac-cess overhead is a novel PHY/MAC design, WiFi-
small as 800 ns.37In WiFi- -Nano (Section 3), all transmitters speculatively trans-mit their preambles in ... is 2.8X the transmission time of a1500 byte packet. In WiFi- -Nano, instead of waiting for SIFS be-fore transmitting the ack ... receiversimply aborts its ack transmission. This speculative ack preambletransmission allows WiFi- -Nano to eliminate SIFS and, thus, reducethe ack overhead.In order ... the packet. Sec-ond, since backoff counter collisions are unlikely in WiFi- -Nano,unfairness caused due to capture-effect between near-and-far ter-minals [8] is ... as extensive simula-tions (Section 6), we demonstrate the effectiveness of WiFi- -Nanoin improving the efficiency of WiFi.The realization of WiFi- -Nano requires analogue self-interferencecancellation [14] and the abilty to transmit ... latter requires an extra oscillator and antenna. Another over-head that WiFi- -Nano imposes is the need for a longer preambledetection time ... is the need for a longer preambledetection time compared to WiFi. . This is necessary to enable detec-tion of the preamble ... transmissions. Surprisingly, we find that the pream-ble detection time for WiFi- -Nano turns out to be only about 4 slonger than ... turns out to be only about 4 slonger than in WiFi. . The reason for this small overhead is due to ... quickly reducing overall interference. Thus,we believe that the costs of WiFi- -Nano are small compared to itsbenefits.In summary, we make the ... we make the following contributions:? The design and implementation of WiFi- -Nano, a novel PHY/MACdesign that improves the throughput of WiFi by up to 100% us-ing the following techniques.? The use ... in 802.11 at 600 Mbps2. MOTIVATIONIn this section we motivate WiFi- -Nano by analyzing the over-heads present in 802.11. We start ...
... reduce channel access and collision overheadsand improve the efficiency of WiFi. .3. WIFI- -NANO OVERVIEWFigure 3 depicts the time-line of three WiFi- -Nano transmitters A,B and C operating at 600 Mbps. Comparing ... A,B and C operating at 600 Mbps. Comparing Figures 3 (WiFi- -Nano)and 1 (WiFi) ) there are three key differences.? 800 ns Slots : ... differences.? 800 ns Slots : Instead of using 9 sec slots, WiFi-
... (devices B and C in Figure 3).? Speculative ACK : WiFi- -Nano devices eliminate the need forSIFS by speculatively transmitting the ... errors in the received packet.As a result of these changes, WiFi- -Nano dramatically reduces chan-nel access delays given that they are ... to provide the necessary background, we start by describ-ing why WiFi requires 9 sec slots.Why WiFi uses 9 slotsAs described in 802.11 standard, the need for ... car-rier sensing time Figure 5: Design of slot width in WiFi- -Nanotransmission was initiated in the past slot. A failure to ... the farthest nodes inthe network.Rx-Tx Switching Time : Given that WiFi devices are not requiredto allow simultaneous transmission and reception, several ...
... the carrier sensingduration.In order to enable carrier sensing while transmitting preambles,WiFi- -Nano leverages self-interference cancellation [14]. To avoidthe device?s transmissions from ... due to the devices? own transmissions. To increase robustnessof detection, WiFi- -Nano allows a longer carrier sensing time with-out affecting slot ... trans-mission. Hence, in order to preserve fair access, aborting devicesin WiFi- -Nano roll back their backoff counters to time of initiation ... will be described in greater detailin Section 4.How small can WiFi- -Nano Slots be?Requiring the devices to perform carrier sensing while ... The Near Far ProblemFigure 7: Chained Contention Resolution Example in WiFi- -Nanocorrect estimation of this time. If slot durations are less ... ns), we use a slot width of 800 ns in WiFi- -Nano. Notethat, even with 800 ns slots, WiFi- -Nano will continue to work fornetworks with larger ranges, albeit ... albeit with some unfairness to nodesfarther away.Chained Contention Resolution in WiFi- -NanoConsider an example with three devices A, B and C ...
... in Section 6, the inherentparallelism of chained contention resolution in WiFi- -Nano is ex-tremely quick to resolve all contentions in the ... all contentions in the network. This how-ever, also means that WiFi- -Nano may require larger preambles thanWiFi.Speculative ACKWiFi-Nano also eliminates the ... delay, and time to switchfrom receive to transmit mode. Since WiFi- -Nano nodes have sepa-rate transmit and receive paths, the receiver ... we simplyoverlap the processing of already received data during transmis-sion.4. WIFI- -NANO DESIGN DETAILSIn this section we discuss the various components ... we discuss the various components that allowedfor the implementation of WiFi- -Nano.41As described in Section 3, for chained contention resolution, de-vices ... must abort their transmission as soon as possible. To achievethis, WiFi- -Nano leverages the fact that the stronger the receivedtransmission, the ... able to accurately estimate the time of initiationof the transmission. WiFi- -Nano uses a novel carrier sensing tech-nique - sub-preamble lattice ...
... novellattice correlator as depicted in Figure 9. Each packet of WiFi- -Nano is preceded by a PN sequence comprising several short ... ResolutionSince potential collisions can be detected in each 800ns slot, WiFi- -Nano uses a novel contention resolution scheme to resolve colli-sions ...
... a probability of 1k.Thus, the probabilistic collision resolution mechanism in WiFi- -Nano avoids payload collisions with a high probability, thereby sig-nificantly ... the collision overhead seen in Figure 2.5. TESTBED RESULTSWe implemented WiFi- -Nano on a DSP/FPGA based software de-fined radio platform ? ... platform ? the SFF SDR from Lyretech Inc. Given thatthe WiFi- -Nano MAC is extremely delay sensitive, requiring oper-ations to be ... canceler. In this section we eval-uate three key aspects of WiFi- -Nano ? reliable preamble detection,efficiency and fairness through experiments conducted ... probability 800ns, no self?interference800ns, self?interference4us, self?interfernceFigure 10: Preamble detection in WiFi- -Nano5.1 Reliability of Carrier SensingReliable preamble detection is crucial to ... Carrier SensingReliable preamble detection is crucial to the performance of WiFi- -Nano (Section 3), since missed detections lead to collisions, whilefalse ... = 20.0 us40us20us10us220usFigure 11: Airtime efficiency of WiFiNano5.2 Efficiency of WiFi-
... efficiency (fraction oftime data was transmitted over the air) of WiFi- -Nano. The air-timeefficiency is a function of two key parameters ... air-time efficiency is 17.2%.Note that this is 100% improvement over WiFi. . As the packet sizesare increased to 40 s (3kB packet ... even at larger slots such as9 s and 20 s (corresponding to WiFi slot durations) performanceof WiFi- -Nano is marginally better (2% higher) than that of WiFidue ... 12: Fairness in WiFiNano5.3 FairnessWhen multiple devices share the channel, WiFi- -Nano must allowfair channel access without sacrificing efficiency. Thus, devicesmust ... their counters appro-priately. In order to evaluate fair sharing in WiFi- -Nano, we oper-ated two WiFi- -Nano devices, Node A and Node B, simultaneously.Throughout the experiments, ... Pkts/sec corresponding to an air-time efficiencyof about 17%, corresponding to WiFi- -Nano efficiency as seen inFigure 11. Later, when Node B ... we extend our evaluation using simulations todetermine the scalability of WiFi- -Nano to larger deployments andits robustness to alternative parameter choices. ... the network with highprobability; ii) we evaluate the benefits of WiFi- -Nano in terms ofthroughput and fairness, as compared to 802.11; ... iii) we investigatethe effect of frame aggregation.6.1 Simulation SettingsWe implemented WiFi- -Nano?s preamble detection physical layerin the Qualnet network simulator as ... we use a preamble of8 s for packet detection in WiFi- -Nano, compared to 4 s in WiFi.The reason for this ... hidden nodes. In the presenceof hidden nodes, we believe that WiFi- -Nano would suffer packetcollisions similar to WiFi; ; an approach such as CSMA/CN [17]can be used to ...
... fully backlogged CBR traffic,with packet size of 1480 bytes.By default, WiFi- -Nano reacts to preamble collisions using WiFi?sbinary exponential backoff algorithm. ... However, we also explorethe benefits of using IdleSense [3] for WiFi- -Nano when there ishigh contention. Specifically, we simulate an idealized ... of 350 when there are 30active transmitters.6.2 Preamble Length in WiFi- -NanoWiFi recommends about 4 s of its preamble to perform ... this as the carriersensing preamble length. The rest of the WiFi preamble (36 sin 802.11n) is used for performing other functions ... in order to avoid packet collisionswhile using speculative preamble transmissions, WiFi- -Nano de-vices will require a longer preamble to reliably perform ... require a longer preamble to reliably perform carriersense compared to WiFi devices. Thus, the carrier sensing pream-ble length in WiFi- -Nano has to be longer than that used in WiFi.More ... that used in WiFi.More specifically, the carrier sensing preamble in WiFi- -Nano shouldbe long enough to guarantee that all speculatively transmitting ... is the shortest possible carrier sens-ing preamble 2 length that WiFi- -Nano can use while ensuring thatpreambles are correctly detected even ... 5.1), in noisy environments, higher2Note that the duration of a WiFi- -Nano preamble will be the sumof its carrier sensing preamble ... sensing preamble length and the part of the preambleused in WiFi for other function such as synchronization, channelestimation etc. For example, ... other function such as synchronization, channelestimation etc. For example, a WiFi- -Nano carrier sensing preamblelength of 8 s would result in a ... 802.11a.10 15 200246810121416SINR Detection Threshold in dBPreamble Length in us WiFi? ?Nano with Binary Exponential BackoffWiFi?Nano with IdleSenseFigure 13: Preamble length ... in order to beconservative, we consider an extreme scenario for WiFi- -Nano ? 30full back-logged transmitters, deployed randomly in a 100 ... 30s. We tried two backoffschemes ? exponential backoff (used by WiFi) ) and IdleSense.Figure 13 depicts the minimum carrier sensing preamble ... backoff, a length of 4 s, the same as thatused by WiFi, , is sufficient when the SINR detection threshold is10 dB ... carriersensing preamble length is 7.2 s (an increase of 3.2 s over WiFi) )and 11.2 s (an increase of 7.2 s over WiFi) ) respectively.Also notice that IdleSense has a beneficial effect for ... as to why the preamble length doesnot increase dramatically in WiFi- -Nano, we investigate the pream-ble detection process by evaluating how ...
... morethan 4 s of carrier sensing preamble may be necessary for WiFi- -Nano. Finally, the benefit of IdleSense is evident in this ... the beginning of each epoch is dramat-ically reduced.6.3 Benefits of WiFi- -NanoWiFi-Nano has three main benefits over 802.11. WiFi- -Nano i)significantly reduces the overhead of data transmission, thus in-creasing ... a single transmitter. As the data-rate increases,the relative improvement of WiFi- -Nano over 802.11 increases, dueto high channel access overhead of ... as de-scribed in Section 2. Thus, the throughput gain of WiFi- -Nano over802.11 is 5%, 37%, 85%, 88%, at 6 Mbps, ... 100 200 300 400 500 600020406080100Data?rate in MbpsThroughput in Mbps WiFi? ?Nano 1 nodeWiFi?Nano 30 nodes802.11 1 node802.11 30 nodes 2XGainsFigure ... 30 nodes802.11 1 node802.11 30 nodes 2XGainsFigure 15: ThroughputFigure 16: WiFi- -Nano overheads600 Mbps, respectively. Next, consider the case of 30 ... respectively. Next, consider the case of 30 transmitters.The gap between WiFi and WiFi- -Nano widens further since WiFi- -Nano is able to mostly avoid collisions, while the increased ... channel access overhead. On the other hand, thecollision overhead in WiFi reduces aggregate throughput comparedto the single transmitter case. Thus, at ... the single transmitter case. Thus, at 300 Mbps and 600 Mbps,WiFi- -Nano is able to achieve a throughput gain of 117% ... able to achieve a throughput gain of 117% and 119%over WiFi, , respectively.Similar to Figure 2, we further analyze the performance ... respectively.Similar to Figure 2, we further analyze the performance of WiFi- -Nano by decomposing the overhead into Preamble, ACK + SIFS,Channel ... has been eliminated). Thus, the real over-head that remains in WiFi- -Nano is only the channel access over-head which represents between ... 5.9-16.5% in the Figure.Fairness. Finally, we investigate the benefit of WiFi- -Nano in termsof fairness. In 802.11, the backoff counters of ... closer to an AP which benefit from a higher SINR.In WiFi- -Nano, the preamble detection process terminates with asingle node transmitting ...
... and far nodes. Figure 17 shows theper-node average throughput for WiFi- -Nano and 802.11 at differ-ent combinations of data-rate and number ... and auto-rate fallback, respectively, are usedwhile in the case of WiFi- -Nano, the gap is less than 1% and 5%, re-spectively. ... the gap is less than 1% and 5%, re-spectively. Thus, WiFi- -Nano is able to re-establish fairness withoutpenalizing the total network ... network throughput.6.4 Frame aggregationOne way of reducing the overhead in WiFi is by transmittinglarger packets so that the MAC overhead is ... KB.From the Figure, we see that using 64 KB frames, WiFi effi-ciency approaches 70% and 80% at 600 Mbps and 300 ... re-spectively. Even for the fully aggregated 64 KB frame sizes, WiFi- -Nano is able to achieve 5?10% throughput gains over WiFi. . How-ever, 64 KB average frame sizes are hard to ... 20 30 40 50 60 7000.10.20.30.40.50.60.70.80.91Aggregated frame size in KBEfficiency WiFi? ?Nano 300 Mbps802.11 300 MbpsWiFi?Nano 600 Mbps802.11 600 MbpsFigure 18: ... frame size of 18 KB for large data transfers [18], WiFi- -Nano achieves 25% (resp., 17%) gains over 802.11 at 600 ... above results indicate that frame aggregation can beeffective in reducing WiFi inefficiency for large data transfers, itdoes not help for delay ... voice-over-IPor short HTTP transfers. Thus, we expect higher efficiency gainsfor WiFi- -Nano in typical WiFi settings.7. RELATED WORKThere has been tremendous amount of work targeted ... wireless performance. We discuss a few papers that arerelevant to WiFi- -Nano.Performance. Researchers have proposed several techniques toimprove performance in wireless ... 13], and theuse of aggregation and TDMA-like schedule to improve WiFi ef-ficiency in the presence of VoIP traffic [20]. Perhaps closest ... [18].Idle Sense proposes an alternative to the binary exponential backoff-based WiFi MAC by trying to ensure that hosts in a single ... our evalua-tions, the low collision probability of Idle Sense helps WiFi- -Nanouse a shorter preamble length. FICA tackles the inefficiencies of ... decoding at MIMO data rates over 20-40MHz band-width and at WiFi transmit power is still a challenging, open prob-lem [1]. Finally, ... still suffer from the channel access overhead is-sues of WiFi.In WiFi- -Nano, we only need to correlate with the preamble dur-ing ...
strin-gent as full-duplex systems. Since WiFi- -Nano reduces the channelaccess overhead, it is complementary to full ... is not feasible dur-ing collisions, thereby reducing the collision overhead. WiFi- -Nanoalso suffers from hidden terminal problems like WiFi and couldbenefit from CSMA/CN and ZigZag decoding to improve efficiencyin ... correlating with a preamble-like sequence while packet transmission is on-going. WiFi- -Nano?sspeculative preamble relies on similar correlation ability at the trans-mitter ... layer and above. Thus, instead of 9 s slotsused in WiFi, , we propose WiFi- -Nano that uses slot sizes as smallas 800 ns. The ... a key requirement foraccurate rollback of speculative preamble transmissions.Finally, since WiFi- -Nano nodes are able to detect collisions dur-ing the preamble ... with highprobability. Using testbed experiments and extensive simulations,we show that WiFi- -Nano is able to double the MAC throughput ofWiFi at ...

Title:
BeHop: a testbed for dense WiFi networks

Keywords:
enterprise wifi
dense wifi
residential wifi

Abstract:
<p>We present BeHop, a wireless testbed for dense WiFi networks often seen in residential and enterprise settings. BeHop aims ... on the operation of dense deployments, and evaluate how different WiFi management strategies affect user experience and network behavior. It has ...

References:
A. Patro, S. Govindan, and S. Banerjee. Observing home wireless experience through wifi aps. In Proceedings of the 19th Annual International Conference on Mobile Computing & Networking, MobiCom '13, pages 339--350, New York, NY, USA, 2013. ACM.

Full Text:
BeHop: A Testbed for Dense WiFi NetworksBeHop: A Testbed for Dense WiFi NetworksYiannis Yiakoumis Manu Bansal Adam CovingtonJohan van Reijendam Sachin Katti ... Nick McKeownStanford University{yiannisy,manub,gcoving,jvanreij,skatti,nickm}@stanford.eduABSTRACTWe present BeHop, a wireless testbed for dense WiFi net-works often seen in residential and enterprise settings. Be-Hop aims ... on the operation of dense de-ployments, and evaluate how di?erent WiFi managementstrategies a?ect user experience and network behavior. Ithas su?cient flexibility ... Computer-Communication Networks?GeneralGeneral TermsDesign, Experimentation, Management, Measurement, Per-formance, ReliabilityKeywordsProduction Testbed; Dense WiFi; ; Residential WiFi; ; Enter-prise WiFi; ; Flexibility1. INTRODUCTIONDense WiFi networks ? where multiple APs and clientscoexist ? are commonplace ... arechaotically managed and deployed. A vast majority of usersinstalls their WiFi APs with factory default settings, andlacks the sophistication ? or ... to impractical assumptions or overlooked factors.On the other hand, enterprise WiFi networks apply a com-bination of WiFi management schemes to improve networkperformance, but their platforms are proprietary ... and evaluatingthe pros and cons of new ways of controlling WiFi networks.For example, we plan to evaluate the benefits of centrallycontrolling ... we plan to evaluate the benefits of centrallycontrolling a large WiFi network, controlling power, chan-nel allocation and association from a single ... a single vantage point.To do this, we built a general-purpose WiFi testbed in ouruniversity. We set out with two main requirements ... real users, runningany application they choose on any of their WiFi- -connecteddevices (laptops, phones, tablets, DVRs, etc). This will letus evaluate ...
... BeHop, an operational testbedwhich aims to provide insights on dense WiFi networks, andshare our experiences on how to design and build ... this paper are thefollowing:? We design and prototype an SDN WiFi frameworkwhich can be used to realize a wide set of ... and as-sociation control. We use a virtual-AP abstraction todecouple our WiFi logic from the physical infrastruc-ture and control how to expose ... tousers (sparse or dense, home or enterprise network).We implement per-client WiFi control to simplify man-agement, ease troubleshooting, and apply and evaluatepolicies ... flexibility and control.? We evaluate a novel deployment strategy for WiFi test-beds which practically improves realism and actual us-age. BeHop runs ... have to change any setting touse the network (such as WiFi network preferences).When our network fails or during an upgrade, they ... of dense WiFinetworks and systematically evaluate benefits and implica-tions of WiFi management techniques. We evaluate BeHopthrough one specific usecase serving as ... and evaluation, and iii) a centralized controller wherewe implement our WiFi management logic (Fig. 1(a)).2.1 BeHop APOur dataplane consists of commodity ... consists of commodity APs. Besides packetforwarding, BeHop APs implement low-level WiFi tasks withstrict timing constraints (namely rate adaptation, beacon,and acknowledgement generation). ... a wireless station by adding anentry in the client table.Forward WiFi managment tra?c: BeHop APs forwardWiFi management tra?c to the controller. ... power allocation.Each AP additionally has a monitoring agent (MA) whichforwards WiFi statistics and raw WiFi ... packets to a collec-tor (along with radio metadata for SNR, WiFi rate, andretries). We can configure what statistics to collect (e.g.channel ... which pack-ets to collect (we currently use Probe Requests and WiFi-
... of the network. Our collector hasthree sources of data: i) WiFi statistics reports from APs, ii)raw WiFi packets with radio metadata (SNR, retries, WiFirate), and iii) a ... a 48-bit address.2Internet&CL4&CL5&CL1&CL2&BeHop&AP&VAP1& VAP2&VAP4&CL4&CL3&CL2&CL1&BeHop&AP&Legacy&AP&BeHop&Controller&?& ?&MA&MA&BeHop&Collector&CL5&(a) Overall BeHop architecture. (b) Deployed WiFi APs. 39 out of 150 residents participateon BeHop so far.Figure ... support mobility.Per-client control: Our controller allows us to define thedesired WiFi management strategy on a per-client basis.This is helpful both for ... replicate partsof the functionality.Our datapath consists of commodity dual radio WiFi APsrunning OpenWRT. 2 BeHop APs run Open vSwitch (OVS)along with ... 2 BeHop APs run Open vSwitch (OVS)along with our SDN WiFi extensions that implement thecontrol API. APs connect to the backbone ... build ourcontrol plane using the POX OpenFlow controller [8].To forward WiFi management tra?c, we add a monitorinterface for each radio and ... re-quired significant engineering and debugging e?ort, as cov-ering all possible WiFi capabilities and client diversity isnon-trivial.To activate multiple VAPs on a ... a single AP, we exploitAtheros? BSSID mask. Atheros hardware generates WiFi ac-knowledgements for packets destined to a target that matchesits own ...
... across 8 channels.We extend OVS? JSON-based configuration mechanism(OVSDB) to configure WiFi settings (channel and power),add and remove Virtual-APs and stations. We ... BeHop in a real-world setting is essential tostudy and evaluate WiFi management strategies and theirimplications under the conditions found in a ... want our testbed to carry real tra?c of usersfrom any WiFi- -connected device; at the same time, we wantto maintain the ... not have to manu-ally choose our network, or change their WiFi preferencesand SSID priority-list on their devices (a task which onlybecomes ... we had to trade-o? control and visibility. A client3Our university WiFi network supports campus-wide mo-bility, and often users are already authenticated ...
... is the protocol used between Cisco?s APs andcentralized controller.4Keeping a WiFi testbed active is not trivial. Our previousexperience [14] has taught ... of our deployment sofar with low CPU usage. Forwarding of WiFi managementtra?c to the controller generates a peak-load of 630 packets-per-second. ... to use the 5GHz band to improveperformance on a dense WiFi network.4.1 Usecase: Studying 5GHz Band SteeringBetter use of 5GHz is ... Band SteeringBetter use of 5GHz is a common strategy for WiFi net-works to leverage less interference and wider channels avail-able on ... it hard to provide good coverage. Asa result, we operate WiFi networks on a ?dual-band? modewhere each AP is configured to ... connectivity. Vendors and operators try tofacilitate use of 5GHz through WiFi management featuresoften referred to as band-steering.We use BeHop to evaluate ...
... users and accounts for everything that might degradeperformance in the WiFi infrastructure: bu?er build-up onthe AP due to a link/channel slowdown, ... in poor-coverage for the low-penetration 5GHz-band,and often weak-links and low WiFi rates as clients associatewith APs which are far away. We ...
... growing baseline overhead. There are two main sourcesof overhead in WiFi networks ? WiFi beacons and nor-mal broadcast tra?c (e.g. ARP, mDNS). Each (virtual) ... we looked at Stanford?sComputer Science department, which runs an enterprise-grade WiFi network [5] with 44 APs spread across the spec-trum and ... of tra?c, 2.4GHz approaches maximum utilizationwhile 5GHz remains mostly free.(a) WiFi density at 2GHz. (b) WiFi density at 5GHz.Figure 4: WiFi density at 2.4GHz and 5GHz. Each line represents a link ... be really valuable in exposing and studying such issuesmet in WiFi networks.5. RELATEDWORKManagement of WiFi networks has been widely studied inthe past. MDG [3] combines ...
... deployments inthe recent past. TFA [2] is a wide-area mesh WiFi deploy-ment. TFA?s research focus has been on optimizing topologycontrol. Our ... recent testbeds follow the SDN approach. Open-Roads [14] uses OpenFlow WiFi APs and WiMax base-stations to experiment with flow mobility and ... and interfaces. OpenRoads looks at L2-layer andabove, taking the default WiFi operation for granted. Weare mostly interested in understanding how WiFi manage-7ment a?ects network performance. ODIN [13], a frameworkfor enterprise WiFi, , exposes a BSSID for every client, whichis then treated ... existing networks nothaving the flexibility to host our experiments.Finally, commercial WiFi systems use centralized con-trollers to optimize network management and performance[7, ... early experience with BeHop let us evalu-ate popular algorithms for WiFi, , and revealed trade-o?s thatwere not obvious before. We are ... testbeds have great potential to inform the design ande?ciecy of WiFi networks, and most techniques from BeHopcan be reused to build ... Patro, S. Govindan, and S. Banerjee. Observinghome wireless experience through wifi aps. InProceedings of the 19th Annual InternationalConference on Mobile Computing ...

Title:
The Wireless Shark: Identifying WiFi Devices Based on Probe Fingerprints

Keywords:
wifi scanning
wifi probing patterns

Abstract:
<p>Due to the broadcast nature of WiFi communication anyone with suitable hardware is able to monitor surrounding ... suitable hardware is able to monitor surrounding traffic. However, a WiFi device is able to listen to only one channel at ... given time.</p> <p>In this paper we present an inexpensive multi-channel WiFi capturing system (dubbed the wireless shark") and evaluate its performance ... that the performance is directly related to the number of WiFi adapters being used for listening. As a second contribution of ... that the devices send when they attempt to discover available WiFi networks. Our results expose some distinct characteristics in various mobile ...

Full Text:
The Wireless Shark:Identifying WiFi Devices Based on ProbeFingerprintsOtto WaltariDepartment of Computer ScienceUniversity of Helsinkiotto.waltari@helsinki.fiJussi ... of Computer ScienceUniversity of Helsinkijussi.kangasharju@helsinki.fiABSTRACTDue to the broadcast nature of WiFi communicationanyone with suitable hardware is able to monitor sur-rounding traffic. ... suitable hardware is able to monitor sur-rounding traffic. However, a WiFi device is able to lis-ten to only one channel at ... theintuition that the performance is directly related to thenumber of WiFi adapters being used for listening. As asecond contribution of the ... that the devices send when they attempt to dis-cover available WiFi networks. Our results expose somedistinct characteristics in various mobile devices? ... mobile devices? prob-ing behavior.1. INTRODUCTIONMost of today?s mobile devices have WiFi network-ing capabilities and it is common for users to rely ... better perfor-mance compared to, e.g. 3G or LTE, or because WiFi ischeaper (typically free) compared to cellular networks.Permission to make digital ... . . $15.00DOI: http://dx.doi.org/10.1145/2935755.2935757Some mobile operators are even o?oading to WiFi inorder to reduce traffic on the cellular network. Fur-thermore, some ... the point of refusing to performcertain operations unless connected to WiFi (e.g. iOSdevices and downloading of large files).Because WiFi is a broadcast medium anyone withcommodity hardware is able to ... peri-ods, thus violating privacy of mobile users.Most related work involving WiFi monitoring impliesthat the monitoring system relies on channel hoppingin order ... how easily andcost efficiently we can build a comprehensive multi-channel WiFi monitoring device. Our validation of thewireless shark shows how its ...
... been studied and deployed [4][6].Barbera et al. [3] presents a WiFi monitoring setupthey used to gather probe requests at public gatherings.They ... operationon different frequencies and data rates. Most popu-lar amendments for WiFi devices to utilize are b, gand n. All of them ... order to monitor all channels we decided to ded-icate one WiFi adapter per channel, i.e. a total of13 adapters in our ...
... show this weused three different traffic generators; one generated ar-tificial WiFi probe requests at a controlled rate, anotherone maintained an idle ... truth (i.e. case of 13/13).4. EXPOSING PROBING BEHAVIORNetwork discovery in WiFi can be done in two distinctways; either through active or ... reasons. First of all, a sniffing observer hasto be in WiFi proximity for a long time, which maynot be possible since ... requests are sent in bursts that sweep throughall the channels WiFi bandwidth is divided into. An-other way of classifying devices is ...
... taking a closer look at individualbursts of probe requests. A WiFi network can be con-figured to any of the available channels. ... environments. In these data sets target devicesdo not have known WiFi networks in range. Totalrecording time of the data set is ...
... starting from the be-ginning of a burst, and Y-axis represents WiFi channel(1 through 13). Dots illustrate probe requests sent inall of ...
... for future work.6. CONCLUSIONIn this paper we presented a multi-channel WiFi cap-turing system which we dubbed the ?wireless shark?.We evaluated its ... that our capturing system could bebeneficial in any application where WiFi is being mon-itored and a higher capturability is favorable.We used ... that are sent by devices inorder to cover all available WiFi channels. From thesebursts we could identify clear device specific patternsand ... arguments regarding our findings.As future work we intend to analyze WiFi behaviorof mobile devices further and look at how devices actwhile ...

Title:
Look who's talking: a practical approach for achieving scheduled wifi in a single collision domain

Keywords:
scheduled wifi
centralized wifi

Abstract:
... following question in this paper: <i>Can the goals of centralized WiFi scheduling be achieved using purely distributed operations?</i> We present a ... for arbitrary schedules to be distributed to nodes in a WiFi network. The nodes in the network then use purely local ...

References:
"iPass WiFi Growth Map." {Online}. Available: http://www.ipass.com/wifi-growth-map/

Full Text:
... the following question in this paper: Can the goalsof centralized WiFi scheduling be achieved using purely dis-tributed operations? We present a ... for arbitrary schedules to be dis-tributed to nodes in a WiFi network. The nodes in the net-work then use purely local ... ns3) to evaluate LWT.CCS Concepts?Networks?Network protocols; Wireless local area net-works;Keywordscentralized WiFi, , scheduled WiFi, , media access control1. INTRODUCTIONMost WiFi deployments today use the distributed coordi-nation function (DCF) mode of ... in environments where operationalefficiency is an issue (e.g., in high-density WiFi de-ployments), centralized scheduling can lead to higherefficiencies.On the other hand, ... scalability with the number of nodes, and backwardcompatibility with how WiFi is predominantly used today[4].The context for this paper is this ...
... of LWT, and Section 5concludes the paper.2. BACKGROUND AND PROBLEMDEFINITION2.1 WiFi DCF - A PrimerThe Distributed Coordination Function (DCF) mode ofIEEE ... a DCF transmission Figure 2: System architecture for scheduled WiFiScheduled WiFi is the notion of making WiFi nodes trans-mit according to an order specified in a target ... (which can either be an on-site controller in an enterprise WiFi deployment or a cloudcontroller for APs that do not belong ... the same adminis-trative domain) communicates with all APs in the WiFi net-work. Protocols such as CAPWAP([5]) and LWAPP([6]) fa-cilitate the communication ... LWAPP([6]) fa-cilitate the communication between the APs and the centralcontroller. WiFi- -related network information is sent to thecentral controller periodically. Based ... a metric called ad-herence, adh, to measure how well the WiFi nodes trackthe prescribed schedule. The transmission pattern of WiFinodes is ... a transmission pattern2.3 Related WorksIt is possible to achieve scheduled WiFi using a purelycentralized MAC protocol such as the point coordinationfunction ...
... is the same as that of the 802.11standard. Thus, general WiFi nodes can support LWT?s re-quired time synchronization level.Hybrid MAC protocols ... 4.2.4 Scope and ChallengesThe focus of this paper is on WiFi networks in a singlecollision domain, in which any two simultaneous ... domain, in which any two simultaneous transmis-sions cause a collision. WiFi network deployments typicallyhave auto-channel-selection mechanisms (3 to 12 orthog-onal channels ... by this paper can be statedas follows. Consider a multi-cell WiFi network containingn nodes in a single collision domain. The central ... to all nodes.Having S, how can the nodes achieve scheduled WiFi ef-ficiently? We present below a list of non-trivial challengesthat need ... of non-trivial challengesthat need to be addressed by any scheduled WiFi solution:Non-backlogged Nodes: A common problem in sched-uledWiFi is to deal ... deal with non-backlogged nodes without collectingqueue status.Decodability vs. Detectability: The WiFi PHY layeruses multiple rates for data transmissions. Thus, overheardpackets cannot ... collision domains. A well-known solutionfor the hidden terminal problem in WiFi is the exchange ofRTS/CTS before data transmissions. However, RTS/CTS in-troduce ... this in Section 3.4). Thus, it is desirable to achievescheduled WiFi even in the presence of hidden terminals.Backward Compatibility: Since it ...
... nodes to operate nor-mally without additional overheads and delay.Sleeping Nodes: WiFi radios can be put to sleep to con-serve energy. Nodes ... time elapsed to estimate the current schedule slotafter sleeping. However, WiFi supports multiple rates fordata transmissions, and packet sizes can vary ...
... in LWT-Baseline.3.1.3 Limitations of Baseline algorithmEven though LWT-Baseline achieves scheduled WiFi effi-ciently, it is limited to simple scenarios where nodes are ... LWT(LWT-WC), which utilizes the carrier sense/clear channelassessment (CS/CCA) mechanism of WiFi to automaticallyrelease unused resources to other nodes when the schedulednode ... core insight of LWT-WC lies in the CS/CCA mecha-nism of WiFi. . WiFi packets contain a PLCP preamble in thebeginning for synchronization. It ...
... becomes a rand slot.3.2.3 Experimental Validation of Work-ConservingWe set up WiFi communications using WARP [12] to val-idate the core insight of ...
... signals can use any of the 36 sub-carriers in a WiFi channel, there are at least 36 different flashsignals, and transparent ... second.Switch to Transmit while ReceivingDue to the half-duplex property of WiFi radios, a node cannot transmit a flash signal while receiving. ...
... updating S.3.5.4 Multiple collision domainsSo far, we propose algorithms for WiFi networks in a sin-gle collision domain. It is possible to ... MAC and PHY layer, and can operate fastenough to perform WiFi communications with off-the-shelfWiFi devices.As shown in Fig. 12, we set ...
... A SoftwareTDMA-based MAC over Commodity 802.11 hardware,?IEEE INFOCOM, 2009.[4] ?iPass WiFi Growth Map.? [Online]. Available:http://www.ipass.com/wifi-growth-map/[5] B. O?Hara and P. J. K. ...

Title:
Sparse WiFi deployment for vehicular internet access with bounded interconnection gap

Keywords:
roadside WiFi

Abstract:
<p>Vehicular Internet access via open WiFi access points (APs) has been demonstrated to be a feasible ...

References:
A. Balasubramanian, R. Mahajan, A. Venkataramani, B.N. Levine, and J. Zahorjan, "Interactive WiFi connectivity for moving vehicles," in Proc. ACM SIGCOMM, Sep. 2008, pp. 427-438.
J. Eriksson, H. Balakrishnan, and S. Madden, "Cabernet: Vehicular content delivery using WiFi," in Proc. ACM MobiCom, Sep. 2008, pp. 199-210.

Full Text:
... IEEE/ACM TRANSACTIONS ON NETWORKING, VOL. 20, NO. 3, JUNE 2012Sparse WiFi Deployment for Vehicular InternetAccess With Bounded Interconnection GapZizhan Zheng, Student ... IEEE, and Santosh Kumar, Member, IEEEAbstract?Vehicular Internet access via open WiFi accesspoints (APs) has been demonstrated to be a feasible solution ... is demonstrated via simulations using data fromreal-world road networks.Index Terms?Roadside WiFi, , sparse coverage.I. INTRODUCTIONT HE GROWING popularity of media-enabled handheldssuch ...
... interconnection gap as the1063-6692/$26.00 2011 IEEEZHENG et al.: SPARSE WIFI DEPLOYMENT FOR VEHICULAR INTERNET ACCESS WITH BOUNDED INTERCONNECTION GAP 957maximum ...
... best of our knowledge, the deployment issues with re-spect to WiFi- -based Vehicular Internet Access have not beencarefully studied so far, ... We are now ready to formally de-ZHENG et al.: SPARSE WIFI DEPLOYMENT FOR VEHICULAR INTERNET ACCESS WITH BOUNDED INTERCONNECTION GAP 959fine ...
... small number of equivalence classes, and itZHENG et al.: SPARSE WIFI DEPLOYMENT FOR VEHICULAR INTERNET ACCESS WITH BOUNDED INTERCONNECTION GAP 961suffices ...
... distance label. For each vertex and eachZHENG et al.: SPARSE WIFI DEPLOYMENT FOR VEHICULAR INTERNET ACCESS WITH BOUNDED INTERCONNECTION GAP 963terminal ...
... general, one needs to consider all theZHENG et al.: SPARSE WIFI DEPLOYMENT FOR VEHICULAR INTERNET ACCESS WITH BOUNDED INTERCONNECTION GAP 965Fig. ...
... length of the path is at leastZHENG et al.: SPARSE WIFI DEPLOYMENT FOR VEHICULAR INTERNET ACCESS WITH BOUNDED INTERCONNECTION GAP 967Fig. ...
... Balasubramanian, R.Mahajan, A. Venkataramani, B. N. Levine, andJ. Zahorjan, ?Interactive WiFi connectivity for moving vehicles,? inProc. ACM SIGCOMM, Sep. 2008, pp. ... Eriksson, H. Balakrishnan, and S. Madden, ?Cabernet: Vehicularcontent delivery using WiFi, ,? in Proc. ACM MobiCom, Sep. 2008, pp.199?210.[8] J. Ott ...
... 24, no. 11, pp. 2127?2136, Nov. 2006.ZHENG et al.: SPARSE WIFI DEPLOYMENT FOR VEHICULAR INTERNET ACCESS WITH BOUNDED INTERCONNECTION GAP 969[22] ...

Title:
WiFi centric power modeling of smartphones

Keywords:
WiFi

Full Text:
WiFi centric power modeling of smartphonesWiFi Centric Power Modeling of SmartphonesJonghoe ... of System]: [Modeling Techniques]KeywordsWiFi; Smartphone; Power modeling1. MOTIVATIONPrevious work of WiFi power modeling proposes the power-throughput curve [2], the power-transmission/reception air-time ... which takes into account the power saving opera-tions of the WiFi module so that it accurately estimatespower level of the smartphone ... of the smartphone when the smartphone trans-mits/receives packets using its WiFi interface. Furthermore,we manage the power state of the smartphone with ... mWInactivity time 210 msInactivity time 210 msPower (mW)Time (ms)WiFi module (channel sensing)+ CPUWiFi module (channel sensing)+ CPU WiFi PSMPacket Rx/TxNo packets during inactivity timeNo packets during inactivity timeWiFi ... CPU utilizationWiFi Multicast RxWiFi Tx, RxPWiFi_promWiFi Tx, RxPWiFi_promin < ttailDisplay OnPDisplay_prom???WiFi Multicast Rx?Display OnPDisplay_promDisplay OffPDisplay_tailin < ttail?in > ttail?in > ttail?Display ... Rx?Display OnPDisplay_promDisplay OffPDisplay_tailin < ttail?in > ttail?in > ttail?Display On, WiFi Tx,RxPDisplay,WiFi_promFigure 2: State transition diagram.where in is the packet interval, ... in, u( ) is the unit step function,and ttail is the WiFi inactivity timer timeout value of thesmartphone. P (in) is the ... only up tottail after which the smartphone turns off its WiFi moduleto enter the power save mode (PSM). It consists of ...
power atthe WiFi module and transmission/reception power of thepacket at the WiFi module. The coefficients of Eq. (2), ?and ?, are determined ... power save mode, respectively.Packet interval which is larger than the WiFi inactivitytimer timeout value, ttail, makes the WiFi module turn offuntil the next packet to be transmitted/received is ... power model considers the power reduc-tion by turning off the WiFi module when the smartphoneoperates on the PSM. The FSM also ... and de-lay for promotion whereas the tail power of the WiFi moduleis negligibly small. The display also causes both the promo-tion ...

Title:
A Little Sharing Goes a Long Way: The Case for Reciprocal Wifi Sharing

Keywords:
reciprocal wifi sharing

Abstract:
<p>Widespread deployment of private home Wifi access points (APs) can result in uncoordinated and overlapping wireless ... the world's population. Broadband penetration and the demand for high-speed Wifi throughout at home will lead to more private APs, which ... better use of existing private home APs. We define reciprocal Wifi sharing as cases where two users both improve their network ... their network performance by connecting to each other's overlapping private Wifi networks. Compared to previous approaches that attempted to use private ... that attempted to use private APs to create large-scale open-access Wifi networks, reciprocal Wifi sharing relationships more closely mirror existing human relationships and can ... without elaborate reputation mechanisms.</p> <p>To evaluate the potential for reciprocal Wifi sharing, we analyze 21 M Wifi scans collected from 254 smartphones over 5 months. Our results ... results show that even in a sparsely-populated suburban area, reciprocal Wifi ... sharing can be beneficial. And surprisingly, we detected several reciprocal Wifi sharing opportunities even within our tiny user sample. Motivated by ... we present the design of WiseFi, a system enabling reciprocal Wifi

Full Text:
... A Little Sharing Goes a Long Way:The Case for Reciprocal Wifi ... SharingA Little Sharing Goes a Long Way:The Case for Reciprocal Wifi SharingJinghao Shi, Liwen Gui, Dimitrios KoutsonikolasChunming Qiao and Geoffrey ChallenDept. ... Science and Engineering, University at Buffalo{jinghaos,liwengui,dimitrio,qiao,challen}@buffalo.eduABSTRACTWidespread deployment of private home Wifi access points (APs)can result in uncoordinated and overlapping wireless networks ... investigate whether we can prevent this viciouscycle by using reciprocal Wifi sharing to make better use of exist-ing private home APs. ... better use of exist-ing private home APs. We define reciprocal Wifi sharing as caseswhere two users both improve their network performance ... their network performance by con-necting to each other?s overlapping private Wifi networks. Com-pared to previous approaches that attempted to use private ... approaches that attempted to use private APs tocreate large-scale open-access Wifi networks, reciprocal Wifi shar-ing relationships more closely mirror existing human relationshipsand can be ... maintained without elaborate reputation mechanisms.To evaluate the potential for reciprocal Wifi sharing, we analyze21 M Wifi scans collected from 254 smartphones over 5 months.Our results show ... months.Our results show that even in a sparsely-populated suburban area,reciprocal Wifi ... sharing can be beneficial. And surprisingly, we de-tected several reciprocal Wifi sharing opportunities even within ourtiny user sample. Motivated by these ... results, we present the designof WISEFI, a system enabling reciprocal Wifi sharing.1. INTRODUCTIONTwo trends are combined to create increasingly crowded and ... trends are combined to create increasingly crowded and un-coordinated home Wifi environments. First, increasing broadbandpenetration is creating larger numbers of private ... end of 2014, 451 Mhouseholds worldwide (25%) would have home Wifi and that thisnumber will continue to grow [7]. Second, an ... ACM. ISBN 978-1-4503-3699-4/15/09 ...$15.00.DOI: http://dx.doi.org/10.1145/2799650.2799652 .Bob's ApartmentAlice's ApartmentFigure 1: Example of Reciprocal Wifi Sharing. Solid arrows representweak connections, while dashed lines represent strong ... transmission powers. We refer to this mutually-beneficial ar-rangement as reciprocal Wifi sharing.Reciprocal Wifi sharing has benefits compared to attempts to useprivate APs to ... establish community networks such as FON [1] orOpenWireless [2]. Reciprocal Wifi sharing opportunities are morelikely to align with existing human relationships, ...
... 5Scans 21,192,417Observed APs 1,197,522Used APs 15,668Wifi Sessions 466,032Table 1: PHONELAB Wifi Dataset Summary. Used APs refers to thesubset of total APs ... the devices participating in the study.But how often is reciprocal Wifi sharing beneficial and possi-ble in practice? To explore these questions, ... analyzing a dataset collected on the PHONELAB smartphonetestbed containing 21,192,417 Wifi scan results from 254 smart-phones over 5 months. Despite the ... only several hundred userswe were still able to observe reciprocal Wifi sharing opportunitiesin our tiny sample. Motivated by these results Section ... a system addressing the practical challengesof establishing and monitoring reciprocal Wifi sharing agreements.We conclude by identifying some open challenges in implementingsuch ... reciprocal sharing opportunity in real life scenar-ios, we obtained a Wifi scan result dataset from PHONELAB1 ( 2.1).We first discuss some heuristics ... explore the re-ciprocal sharing relationships in the dataset ( 2.4).2.1 PhoneLab Wifi DatasetPHONELAB[5] is a public smartphone platform testbed operatedat the University ... In par-ticular, the smartphone platform was modified to log each Wifi scanresult and Wifi connection events naturally generated by the An-droid system. Note that ... same informationcan also be logged by applications with appropriate permissions.A Wifi scan result represents the device?s network visibility, andconsists of multiple ... device?s network visibility, andconsists of multiple entries?each corresponds to one Wifi AP thedevice observed. The content of one entry includes: (1) ... was performed is also logged. Ta-ble 1 summarizes the PHONELAB Wifi dataset.2.2 Home AP DetectionWe focus on home Wifi networks which are more likely to revealstable and immediate reciprocal ... identify the home AP for a device, we look at Wifi sessions thathappened during 12 AM and 4 AM and count ... multiple devices may be as-sociated with the same home AP.2.3 Wifi Session Signal StrengthAfter identifying the home AP for each device, ... first question, we inspect scan results that are re-ported during Wifi sessions with home APs. For each such scanresult, we identify ...
... dataset. Such casesindicate that the device could potentially improve its Wifi perfor-mance by connecting to a neighbor AP which has a ... is well un-derstood that RSSI does not directly translate to Wifi performance,which we will discuss in Section 3.3. Also note that ... the second condition.We classify all scan results reported during home Wifi sessionsinto two categories: sub-optimal and the rest. For each device, ... the time, suggesting that these usersmay benefit from sharing the Wifi access of neighbor APs.Next, we want to answer the question ... the need for a system to detectand enable such reciprocal Wifi sharing.3. SYSTEM DESIGNInspired by the results of the investigation in ... system called WISEFI to detect reciprocal sharing opportu-nities ( 3.1), enable Wifi sharing ( 3.2) and monitor the Wifi per-formance to ensure the sharing remains reciprocal ( 3.3). Figure 5shows ...
... home AP of the device, and neighbor APs?signal strength during Wifi sessions with the home AP. A smart-phone application can be ... directly by user. Once the home AP in-formation is identified, Wifi scan results during sessions with homeAPs can then be logged ... Reciprocal sharing opportunities are detected by WISEFI smartphone apps; (2) Wifi sharing is enabled throughcoordination of WISEFI server; (3) Wifi usage and performance are monitored by WISEFI app to ensure ... wireless routers support the virtual net-work feature, where multiple virtual Wifi networks are emulatedby a single router hardware, and different network ... network. This feature is typically used toset up a guest Wifi network to provide network access to temporalvisitors yet isolate them ... isolate them from home clients. For home APs withsuch feature, Wifi sharing can be achieved by only distributing thecredentials of guest ... complicated for average users to perform. How-ever, simply sharing the Wifi credentials of user?s home AP to otherWISEFI users is not ... be forced to change the home AP password and reconfigurethe Wifi credentials on all his/her devices just to revoke the accessof ... home AP. Evenif the system can directly share each other?s Wifi credentials, man-ually configuring it on all devices is still tedious.To ... is still tedious.To overcome this challenge, we propose a dynamic Wifi AP con-figuration API with two simple interfaces: getAuthClientsand setWhiteList. The ... associated with the AP through normal authen-tication. In the home Wifi network scenario, this interface shallreturn only the MAC addresses of ... interface shallreturn only the MAC addresses of the user?s own Wifi devices.On other hand, setWhiteList sets a list of white list ... in mostcommodity APs.With the help of these configuration APIs, the Wifi sharing pro-cess can work as follows. Suppose the WISEFI system ...
... approach. First, notethat throughout the grant and revoke process, the Wifi credentialsof Alice?s home AP are not shared with Bob or ... users if needed. Finally, this mechanism does notrequire modifications of Wifi clients (except for installation of theWISEFI app) and only requires ... thesharing. For instance, suppose after the system has established re-ciprocal Wifi sharing between Alice and Bob, and Bob decides todeploy an ... so obvious reason is that, as mentioned in Sec-tion 2.3, Wifi signal strength is used as a hint to identify poten-tially ... is well known that signal strength doesnot directly translate to Wifi performance. Other factors, suchas AP load, PHY data rate, interference, ... Other factors, suchas AP load, PHY data rate, interference, or Wifi generation (e.g.,802.11/g/n/ac), also affect the link quality yet can not ... not be easily de-tected by the smartphone. Furthermore, last hop Wifi link qualitydoes not necessarily determine the clients? overall end-to-end net-work ... sharingusage of other WISEFI users to ensure reciprocity.4. OPEN QUESTIONSEnabling Wifi sharing between neighbors both touches knownopen issues of cooperative Wifi access and brings new challenges.As discussed in Section 3.2, user?s ... the peers who share the network.Another challenge in establishing reciprocal Wifi sharing is thebootstrap process. It is expected that during early ...
... to the WISEFI app is to help the user findbetter Wifi channels for their own APs. Uses who are willing to ... a community effort for ubiqui-tous Internet access. Volunteers configure their Wifi network withopen access and a special SSID, openwireless.org, to adver-tise ... activity among allother users who share access to the open Wifi network. On otherhand, FON [1] is a commercial Wifi sharing network, where regis-tered users can roam over FON-supported Wifi networks. WLANowners share their Wifi network either for small money compensa-tion, or to get Wifi access to other users when they are way fromhome (roaming). ... are way fromhome (roaming). FON aims at providing a global Wifi sharingcommunity where users want to connect to others? Wifi networkbecause they are away from home and have no WLAN ... have no WLAN access.Both OpenWireless and FON aim at sharing Wifi access betweenstrangers either through volunteering or financial incentives. Incontrast, in ... volunteering or financial incentives. Incontrast, in our proposal, users share Wifi network locally (withinneighbors) for better network performance, and the sharing ... and stable (physicalneighbor relationship).There are also several works on cooperative Wifi sharing. Di-mopoulos et al. [4] propose a reciprocal Wifi ... sharing mechanismand later extend it to a large scale peer-to-peer Wifi roaming frame-work [3]. They mostly focus on the reciprocal manner ... relationship. Thesereceipts can later on be consumed to get reciprocal Wifi access fromother users. Such reputation mechanisms can also be applied ... explore the reciprocal sharing opportunitiesthrough extensive analysis of the PHONELAB Wifi dataset, andshow that such opportunity does exist despite the spatial ... that detects reciprocal sharing opportunities, enable Wifisharing and monitor the Wifi usage and performance to ensure thesharing remains reciprocal.AcknowledgmentsWe thank the ... Analytics. Global broadband and wlan (wi-fi) networked householdsforecast 2009-2018. https://goo.gl/IUKVfD.10IntroductionInvestigationPhoneLab Wifi DatasetHome AP DetectionWifi Session Signal StrengthReciprocal Sharing OpportunitiesSystem DesignDetectionSharingMonitoringOpen QuestionsRelated ...

Title:
Dynamic clustering in WiFi direct technology

Keywords:
WiFi direct

Abstract:
<p>WiFi Direct is a new technology supported by WiFi Alliance. Devices can establish connections using an Access Point (network ... this paper, we propose a dynamic election of leaders for Wifi Direct Technology. In our approach, devices are exposing clustering service ...

References:
D Camps-Mur; A Garcia-Saavedra; P Serrano, Device to device communications with WiFi Direct: overview and experimentation, IEEE Wireless Communications Magazine, 2012.

Full Text:
Dynamic Clustering in WiFi Direct Technology ?Urbano Botrel MenegatoiMobilisDECOM / UFOPOuro Preto, MG - ... MG - Brasilrrabelo@gmail.comABSTRACTWiFi Direct is a new technology supported by WiFi Al-liance. Devices can establish connections using an AccessPoint (network leader), ... In this paper, wepropose a dynamic election of leaders for Wifi Direct Tech-nology. In our approach, devices are exposing clustering ser-vice ... provides three main contribu-tions:1. Introduce the concept of clusterization in WiFi Directdevices networks;2. Present a template to help future implementations ofclustering ... that provides an ab-straction of the complexity of access to WiFi Directfeatures.By experiments, we show that our proposal is a feasible ... problemsof the technology that can be explored in future years.252. WIFI PEER TO PEERThe growth of mobile and portable devices, such ... communication between two devices with-out notifications or agreements between users. WiFi P2Pdoes not work the same way as the WiFi Ad Hoc modedoes. WiFi P2P implements a peer-to-peer communicationover 802.11 infra-structure mode, inheriting all ...
... each one is well defined and static throughout thecommunication. In WiFi P2P, these roles are dynamicallyassigned and any of the devices ... act simultaneously as both clientand Access Point.2.1 ArchitectureCommunication devices in WiFi ... P2P are called P2P De-vices. So that communication can occur, WiFi P2P devicesestablish logical groups called P2P groups. These groupsare functionally ... groups called P2P groups. These groupsare functionally equivalent to traditional WiFi networks; thedevice that implements the AP (Access Point) function iscalled ... of IEEE 802.11 [2]. Inthis phase, P2P device discovers legacy WiFi networksand P2P GO.2. Find - Aims to guarantee that two ... de-cided in a previous negotiation.2. Provision of P2P group.(a) Use WiFi Simple Configuration (WSC) for ex-changing credentials(b) Establish P2P group with ... and definition of roles, the devices establisha safe connection using WiFi Simple Configuration (WSC)and finally, we use the DHCP for the ... meet and connect to the group using traditionalmethods for discovering WiFi and go straight to the Pro-vision phase and DHCP. Since ...
... can beaddressed by re-affiliation or by restarting the formation pro-cess.4. WIFI DIRECT IMPLEMENTATIONThe core of our implementation is changing the usage ... must usually be implemented takinginto account only the peculiarities of WiFi Direct technol-ogy. Given this scenario, we created a ?template? for ...
... changed. If it isnecessary to create or delete a group, WiFi- -Direct API mustbe used.4.3 ExperimentsTo validate the architecture and test ... ExperimentsTo validate the architecture and test the clustering tem-plate via WiFi Direct, we used the algorithms ?Major ID?,?GEDIR? and ?MCFA? for ...
... and underestimated.During the project and implementation of our proposal forusing WiFi Direct, we observed several important aspects ofthe technology. Among these ... technology. Among these aspects we can mention thelittle use of WiFi Direct and its potential to be explored.We believe our work ... such as mobile computing, distributedsystems, and other similar areas. As WiFi Direct technologyhas been still little explored, there are several opportunitiesfor ... D Camps-Mur; A Garcia-Saavedra; P Serrano, Deviceto device communications with WiFi Direct: overviewand experimentation, IEEE Wireless CommunicationsMagazine, 2012.[2] Wi-Fi Alliance, P2P ...

Title:
HitchHike: Practical Backscatter Using Commodity WiFi

Keywords:
WiFi

Abstract:
... power backscatter system that can be deployed entirely using commodity WiFi infrastructure. With HitchHike, a low power tag reflects existing 802.11b ... low power tag reflects existing 802.11b transmissions from a commodity WiFi transmitter, and the backscattered signals can then be decoded as ... the backscattered signals can then be decoded as a standard WiFi packet by a commodity 802.11b receiver. Hitch-Hike's key invention is ... for widespread deployment of low-power backscatter communication using widely available WiFi infrastructure. We show experimentally that HitchHike can achieve an uplink ...

References:
D. Bharadia, K. R. Joshi, M. Kotaru, and S. Katti. Backfi: High throughput wifi backscatter. In Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication, pages 283--296. ACM, 2015.
P. ZHANG, D. Bharadia, K. Joshi, and S. Katti. Enabling backscatter communication among commodity wifi radios. In Proceedings of the 2016 Conference on ACM SIGCOMM 2016 Conference, SIGCOMM '16, pages 611--612, New York, NY, USA, 2016. ACM.

Full Text:
... low power backscatter system thatcan be deployed entirely using commodity WiFi infrastruc-ture. With HitchHike, a low power tag reflects existing802.11b transmissions ... a low power tag reflects existing802.11b transmissions from a commodity WiFi transmitter,and the backscattered signals can then be decoded as a ... the backscattered signals can then be decoded as a stan-dard WiFi packet by a commodity 802.11b receiver. Hitch-Hike?s key invention is ... doorsfor widespread deployment of low-power backscatter com-munication using widely available WiFi infrastructure. Weshow experimentally that HitchHike can achieve an uplinkthroughput of ... it achieves a throughput of around200Kbps.CCS Concepts?Networks?Network architectures; Wireless accessnetworks;KeywordsBackscatter; WiFi; ; Wireless1. INTRODUCTIONBackscatter communication has recently attracted inter-est for applications ... as to decode the backscat-tered signals. Recent research such as WiFi backscatter [18],BackFi [2] and Passive WiFi [20] has reduced the need forspecialized hardware. Passive WiFi for example can decodebackscattered signal using standard WiFi radios. However,it still requires a dedicated continuous wave signal generatoras ... source. BackFi needs a propri-etary full duplex hardware add-on to WiFi radios to enablebackscatter communication. Inter-Technology Backscatter[14] is a system that ... enables backscatter communicationfrom a commercial Bluetooth radio to a commercial WiFi ra-dio. Despite its novelty, it does not enable backscatter com-munication ... Despite its novelty, it does not enable backscatter com-munication among WiFi ... radios. Consequently, a backscattersystem that can be deployed using commodity WiFi radioson access points, smartphones, watches and tablets, doesnot exist.Further, recent ... watches and tablets, doesnot exist.Further, recent work such as Passive WiFi also requirestwice the spectrum compared to other approaches such asBackFi, ... generate the RF excitation tone signalon the center of two WiFi channels and the backscatter tagfrequency shifts and reflects the excitation ... tagfrequency shifts and reflects the excitation signal on boththe two WiFi channels. Due to the frequency shifting, theWiFi device (the reader) ... a single backscatter communication link,we end up using two 20Mhz WiFi channels. For example,the continuous wave signal generator sends out a ... backscattered signal. We observethat most users are surrounded by multiple WiFi radios, ei-ther in APs or on their phones and tablets ... and tablets or even on theirsmartwatches. HitchHike utilizes these commodity WiFi
... since itcan benefit from the ubiquity and low-cost nature of WiFi- -600-400-200 0 200 400 600 0 100 200 300 400 ... (dBm)Figure 1: HitchHike concept: HitchHike enables backscattercommunication between commodity 802.11b WiFi radios.and eliminate the need for a specialized, dedicated readeror receiver. ... Second, HitchHike does not waste spectrum, itpiggybacks backscattered signals on WiFi packets that arebeing used for productive communication. Hence, HitchHikecan be ... for productive communication. Hence, HitchHikecan be efficiently deployed with current WiFi infrastructureand unlicensed spectrum.HitchHike?s deployment is best explained via the examplein ... Figure 1. The excitation device is a smartphone with astandard WiFi radio. The smartphone transmits an 802.11bpacket to the first AP ... is connected on Channel 1.To backscatter, the tag receives the WiFi packet, frequencyshifts it to Channel 6, modulates its information and ... it to Channel 6, modulates its information and thenreflects the WiFi signal. The second AP, which is tuned tolisten on Channel ... then receives and decodes the backscat-ter packet as a standard WiFi packet. As shown in Figure 1,deploying HitchHike is very simple. ... spectrum. BothWiFi channels are used for productive communication, onefor standard WiFi and the other for the backscatter link.The key challenge in ... in realizing HitchHike is: how can abackscatter tag produce a WiFi compliant packet by backscat-tering another WiFi compliant packet and also modulate itsdata on the resulting packet? ... packet? HitchHike?s key conceptualcontribution here is codeword translation. Specifically,every 802.11b WiFi packet is a sequence of codewords thatare picked from a ... waste spectrum un-like prior work based on frequency shifting. Both WiFi chan-nels are used for productive communication, and no spurioussidebands are ... on the uplink, but note that prior work has alreadydemonstrated WiFi backscatter designs (which can be usedwith HitchHike too) for the ...
... (LOS)deployment, 1.5 longer than the maximum range re-ported by Passive WiFi. . In non-line-of-sight (NLOS)deployment, HitchHike is able to decode the ... that our system is able to co-exist gracefully with existing WiFi infrastructure.2. 802.11B PRIMERHitchHike backscatters 802.11b packets from commod-ity WiFi devices. Here we give a brief description of how802.11b packets ... = barker0 ej?0...code15 = barker15 ej?15(3)In summary, 802.11b WiFi protocol uses a finite set ofcodewords to encode packets. Our ... DESIGNFigure 2 shows an overview of our system. A commod-ity WiFi radio transmits a normal 802.11b WiFi packet,the backscatter tag reflects the packet to another 802.11bWiFi radio ... shifts the frequency of the re-flected signal to an adjacent WiFi channel. The 802.11breceiver listening on the adjacent WiFi channel receives thereflected WiFi packet, decodes the packet using the normalWiFi decoding chain, and ... code0/code1802.11b TX 802.11b RXbackscattered signal802.11b signalself-interference802.11b TX 802.11b RXWiFi packetreflected WiFi packetFigure 2: 802.11b code word translator at the tag.bility, including ... from a received codeword to the actual bit.For a commodity WiFi receiver to decode the backscatteredpacket, its codewords need to come ... need to come from the same code-book as that of WiFi. . In other words, if the backscatter tagcan act as ...
... 802.11b packet (e.g a packet of all 1s)and then any WiFi radio receiving the backscattered packetcan XOR with the known 802.11b ...
... to frequency shift the backscat-tered signal to an adjacent, non-overlapping WiFi channel.However a non-overlapping WiFi channel does not implythat we will not see any self-interference, ... does not implythat we will not see any self-interference, the WiFi trans-mission emits energy in adjacent channels too. Figure 5shows the ... channels too. Figure 5shows the spectral profile of an 802.11b WiFi transmission.It shows that there is still a signal leaking into ...
messages. These messages will re-serve two WiFi channels, one for the normal WiFi transmis-sion and another for backscatter. We do not ask the ... RF envelopTimeline of various events at the tagExcitation signal (802.11b WiFi packet)...P1 P2 P3 Pn-1 Pn...D1 D2 D3 Dn-1 DnFrom the ... Our 802.11b receiver is a MacbookPro laptop, which has a WiFi card that runs the 802.11a/b/g-/n/ac protocols. We use the sniffer ... the backscattered signal. Then,we use tcpdump to analyze the received WiFi packets andextract the backscattered information.We use an Intel 5300 WiFi card on an Intel NUC as thestandard 802.11b transmitter. It ...
... istwice as better compared to prior backscatter systemssuch as Passive WiFi
... also show that our system is able to co-existwith existing WiFi infrastructure well.5.1 HitchHike?s PerformanceFirst, we investigate HitchHike?s range. We quantify ... throughput, BER, and RSSI across distance in line-of-sight deployment where WiFi TX-to-tag distanceis 1m.(BER), and received signal strength indicator (RSSI). Fig-ure ... system is 54m, 1.5 ?2 longer thanthe range reported by Passive WiFi [20] that leverages a sin-gle tone using dedicated hardware as ... achieved is 32m, longer than themaximum distance reported by Passive WiFi [20]. Second,we are able to achieve 100?200kbps throughput when the802.11b ... throughput, BER, and RSSI across distance in non-line-of-sight deployment where WiFi TX-to-tagdistance is 1m.the LOS deployment because there is a wall ... a result, decoding the backscattered signal becomesmuch harder.5.1.3 Impact of WiFi Transmitter PowerWe also evaluate the performance of our system when ...
... the maximum peak power allowedby FCC is 30dBm. However, many WiFi transmitters chooseto send at a lower power level. Figure 14(a) ... Impact of Packet ChecksumWhen a HitchHike tag converts an incoming WiFi packetinto a backscattered packet, it does not know the contentof ... a backscattered packet, it does not know the contentof the WiFi packet. Therefore, although the backscatteredpacket is a valid WiFi packet on the physical layer, its packetchecksum could be wrong. ... long as we configure these receivers into monitor mode.However, some WiFi receivers do not provide this capabilityand they drop packets with ...
... taghas the potential of being deployed without batteries.5.3 Co-existence with WiFi ... NetworksFinally, we investigate how well backscatter communica-tion co-exists with concurrent WiFi communication. In thisexperiment, we deploy a backscatter tag 4m away ... 802.11b transmitter is 3m awayfrom a laptop, which transmits continuous WiFi packets toanother laptop that is 5m away from the the ... 28 30 32 34 36 38CDFThroughput (Mbps)Backscatter OffBackscatter OnFigure 19: WiFi throughput when backscatter is present andabsent.We look at the throughput ... backscatter is present andabsent.We look at the throughput of the WiFi transmission be-tween the two laptops as well as our backscatter ... understand how they impact each other.5.3.1 How does backscatter impact WiFi? ?Figure 19 shows the WiFi throughput between the twolaptops when backscatter is present and when ... the backscatter transmission, the medianWiFi throughput achieved is 33.9Mbps. The WiFi through-put varies between 22Mbps and 38Mbps because of humanmovement nearby. ... below -70dBm, much lower than the signal strengthof the active WiFi ... transmission. As a result, backscattercommunication does not severely impact active WiFi trans-mission.5.3.2 How does WiFi impact backscatter?We now turn to look at the other case ... to look at the other case ? how does ac-tive WiFi transmission between the two laptops impact thethroughput of backscatter? Figure ... thethroughput of backscatter? Figure 20 shows the backscatterthroughput when the WiFi stream between the two laptopsis present and absent. When we ... two laptopsis present and absent. When we turn off the WiFi streamon the laptops, the median backscatter throughput achievedis 236Kbps. When ... the laptops, the median backscatter throughput achievedis 236Kbps. When the WiFi stream is present, the medianbackscatter throughput drops to 220Kbps, 7% ... is small because the backscattered signal is 60MHzaway from the WiFi stream on the frequency domain andany interference leaking in the ... 0.8 1 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35CDFThroughput (Mbps)WiFi OffWiFi OnFigure 20: Backscatter throughput when WiFi is presentand absent.6. RELATED WORKOur system is the first one ... first one that enables backscatter com-munication between two commodity 802.11b WiFi radioswithout any hardware modification. Our work is inspiredby recent progress ... looks at the opportunity ofbackscattering information on top of a WiFi ... signal. [2]leverages full-duplex technique to enable backscatter com-munication with a WiFi AP. [6] emulates BLE transmissionby backscattering a BLE baseband signal ... enables backscatter communication from acommercial Bluetooth radio to a commercial WiFi radio.However, none of these previous work enables backscattercommunication between two ... none of these previous work enables backscattercommunication between two 802.11b WiFi radios withoutany hardware modification or deploying a specific single toneemitter. ... withoutany hardware modification or deploying a specific single toneemitter. Since WiFi radios and access points are ubiquitousand have been integrated in ... in almost every laptop, smart-phone, etc. More importantly, all these WiFi radios andaccess points are backward compatible in terms of beingable ...
... [22] develop a set offull-duplex cancellation techniques to reduce self-interferencein WiFi and MIMO systems. However, we cannot use thesetechniques in our ... words,we allocate another clean channel for backscatter communi-cation.Network coding: The WiFi bits stream XOR used byour system for improving spectrum efficiency ... first backscatter communication systemthat can be deployed completely using commodity WiFi in-frastructure. Further, it is spectrally efficient, unlike priorsolutions it does ... as GSM, 3G and LTE as well as otherflavors of WiFi such as 802.11g/n/ac.8. ACKNOWLEDGMENTSWe thank the shepherd Joshua R. Smith, ... K. R. Joshi, M. Kotaru, and S. Katti.Backfi: High throughput wifi backscatter. InProceedings of the 2015 ACM Conference on SpecialInterest Group ...
... D. Bharadia, K. Joshi, and S. Katti.Enabling backscatter communication amongcommodity wifi radios. In Proceedings of the 2016Conference on ACM SIGCOMM 2016 ...

Title:
Comparing metro-area cellular and WiFi performance: extended abstract

Keywords:
wifi

Abstract:
<p>Cellular and 802.11 WiFi offer two compelling connectivity options for mobile users. The goal ... conditions. To that end, we compare and contrast cellular and Wifi performance using crowd-sourced data from speedtest.net. We consider spatio-temporal performance ... over 15 weeks. In these preliminary results, we find that WiFi performance generally exceeds cellular performance, and that observed characteristics are ...

References:
A. Balasubramanian, R. Majahan, and A. Venkataramani. Augmenting Mobile 3G Using WiFi. In Proceedings of ACM MobiSys '10, San Francisco, CA, June 2010.
V. Birk, S. Rayanchu, S. Saha, S. sen, V. Shrivastava, and S. Banerjee. A Measurement Study of a Commercial-grade Urban WiFi Mesh. In Proceedings of ACM Internet Measurement Conference, Vouliagmeni, Greece, October 2008.
P. Deshpande, X. Hou, and S. Das. Performance Comparison of 3G and Metro-Scale WiFi for Vehicular Network Access. In Proceedings of ACM Internet Measurement Conference, Melbourne, Australia, November 2010.

Full Text:
Comparing metro-area cellular and WiFi performance: extended abstractComparing Metro-Area Cellular and WiFi Performance[Extended Abstract]Joel SommersColgate Universityjsommers@colgate.eduPaul BarfordUniversity of Wisconsinpb@cs.wisc.eduABSTRACTCellular and 802.11 WiFi offer two compelling connectivity op-tions for mobile users. The goal ... conditions. To that end, we compare and con-trast cellular and Wifi performance using crowd-sourced data fromspeedtest.net. We consider spatio-temporal performance as-pects ... collected over 15 weeks. In these preliminary results, we findthat WiFi performance generally exceeds cellular performance, andthat observed characteristics are highly ... C.4 [Performance ofSystems]: Performance attributesGeneral Terms: Experimentation, Measurement, PerformanceKeywords: Cellular, WiFi, , throughput, latency1. INTRODUCTIONThe past five years has witnessed an ... capabil-ities of mobile devices that are both cellular- and 802.11 WiFi- -capable. The combination of a short-range, high-speed capabil-ity, and a ... growing demand fornetwork bandwidth by mobile users.A vexing problem for WiFi- -enabled cell phone users, serviceproviders, and application designers is seeking ... our study is to understand the spatio-temporal characteristics ofperformance of WiFi- -enabled cell phones in a selection of metro ar-eas with ... mobile devices in a metro area. As ex-pected, we find WiFi download and upload performance to be su-perior to cellular performance ... speeds (in kb/s), latency(in milliseconds), and access type (cellular or WiFi)
... dataset, yet one that providesa broad perspective on cellular vs. WiFi performance in metro ar-eas that are diverse in their geographic, ... preliminary results reveal a wide range of characteristics ofcellular and WiFi performance. The raw comparison between thetwo technologies shows that WiFi provides superior download per-formance, with maximum WiFi performance varying widely. Thedifference in upload performance is much smaller, ... the largest metro areas, with performance decreasingfor both cellular and WiFi during the hours of peak use. Com-parisons between metro areas ... area types (notice the different y-axis scales for cellular (top)and WiFi (bottom)). We observe that for cellular access, the perfor-mance for ... has a clear impact on throughput performance. We observe thatfor WiFi connections, while the smallest metro areas have gener-ally lower throughputs, ... the differences are not as great among themetro areas for WiFi as they are for cellular connections.04/2104/2204/2304/2404/2504/2604/2704/2804/290200040006000800010000Average speed per hour (kb/s)newyorkmadisonbrusselsjacksonalmaty04/2104/2204/2304/2404/2504/2604/2704/2804/2905000100001500020000Average ... hour (kb/s)newyorkmadisonbrusselsjacksonalmatyFigure 1: Average hourly performance for cellular downloads(top) and WiFi downloads (bottom) for exemplars in each ofthe metro areas during ... body of work that examines the be-havior and characteristics of WiFi networks. Studies that are mostclosely related to ours have focused ... 8]. While these prior studies expand the bodyof knowledge on WiFi and cellular behavior individually, our workdiffers in objective, scope, measurement ... measurement details, and the fact thatwe include analysis of both WiFi and cellular performance. Lastly,there are several studies that investigate both ... performance. Lastly,there are several studies that investigate both cellular and WiFi per-formance, primarily in vehicular settings [1,3]. Our results comple-ment and ... inmore general (non-vehicular) settings.5. CONCLUSIONS AND FUTURE WORKCellular and 802.11 WiFi are the de facto connectivity optionsfor today?s mobile users. The ...
... consideringrelated datasets such as weather conditions during test periods andcell tower/WiFi access point locations. Finally, we plan to conducttargeted, hypothesis-driven experiments ... A. Balasubramanian, R. Majahan, and A. Venkataramani. Augmenting Mobile3G Using WiFi. . In Proceedings of ACM MobiSys ?10, San Francisco, CA, ... Shrivastava, and S. Banerjee. AMeasurement Study of a Commercial-grade Urban WiFi Mesh. In Proceedingsof ACM Internet Measurement Conference, Vouliagmeni, Greece, October ... X. Hou, and S. Das. Performance Comparison of 3G andMetro-Scale WiFi for Vehicular Network Access. In Proceedings of ACMInternet Measurement Conference, ...

Title:
Understanding wifi-based connectivity from moving vehicles

Keywords:
wifi

Abstract:
... that we have deployed, we analyze the fundamental characteristics of WiFi- -based connectivity between basestations and vehicles in urban settings. Our ...

Full Text:
Understanding WiFi- -based Connectivityfrom Moving VehiclesRatul Mahajan John Zahorjan Brian ZillMicrosoft Research ... that we have deployed, we analyze the fundamental charac-teristics of WiFi- -based connectivity between basestations and vehi-cles in urban settings. Our ... to get brief periods of connectivity [3],these networks and the WiFi technology itself are not designed toprovide connectivity to moving vehicles.Like ... vehicles.Like others before us [10, 7, 3], we ask if WiFi can be leveragedto provide connectivity (in areas of good coverage) ... the vehicle moves.To investigate feasibility, we have deployed a modest-size, WiFi- -based testbed, called VanLAN. It currently has eleven basestationsand two ... vehicles and basestations.We are interested in the basic nature of WiFi- -based connectivity,such as how it varies as the vehicle moves ...
... setup with two BSs along a high-way (with few other WiFi users in the vicinity) [10]. They find thatthe connection between ... the performance of uploading datafrom a moving car using urban WiFi APs that happen to be open [3].It finds that reasonable ... are significant. In contrast, we focus on thefundamental characteristics of WiFi- -based connectivity.Like VanLAN, DieselNet [5] employs a testbed of WiFi- -equippedbuses as well. However, it has a different technical thrust. ...
... quite complex. It is different from that ob-served in previous WiFi- -based studies of controlled environments [10,7]; well-defined phases are absent, ...
... connectivity on average.7. SUMMARYOur work uncovers a complex picture of WiFi-
... gray periods are likely to be part and parcel of WiFi- -based ve-hicular access. Because its difficult to provide blanket good ...

Keywords:
roadside WiFi

Abstract:
... guarantees. In contrast, a carefully planned <i>sparse</i> deployment of roadside WiFi provides an economically scalable infrastructure with quality-of-service assurance to mobile ...

References:
A. Balasubramanian, R. Mahajan, A. Venkataramani, B. N. Levine, and J. Zahorjan, "Interactive WiFi connectivity for moving vehicles," in Proc. ACM SIGCOMM, Sep. 2008, pp. 427--438.
J. Eriksson, H. Balakrishnan, and S. Madden, "Cabernet: vehicular content delivery using WiFi," in Proc. ACM MobiCom, Sep. 2008, pp. 199--210.
Google WiFi, Mountain View, CA, USA, "Google's Mountain View WiFi network," {Online}. Available: http://wifi.google.com/
Z. Zheng, P. Sinha, and S. Kumar, "Sparse WiFi deployment for vehicular Internet access with bounded interconnection gap," IEEE/ACM Trans. Netw., vol. 20, no. 3, pp. 956--969, Jun. 2012.

Full Text:
... performance guarantees. In contrast, a carefully plannedsparse deployment of roadside WiFi provides an economicallyscalable infrastructure with quality-of-service assurance to mobileusers. In ... and average case compared to some commonlyused deployment algorithms.Index Terms?Roadside WiFi, , scalable deployment.I. INTRODUCTIONW ITH increasing popularity of media-enabled handheldsand ... in this paper are available onlineat http://ieeexplore.ieee.org.Digital Object Identi?er 10.1109/TNET.2014.2309991support. WiFi hotspots are rapidly mushrooming in every city tomeet this demand. ...
... data access by mo-bile users using ?in situ? (or ?open?) WiFi networks [7], [10],[12], [22] and in various controlled environments [7], ... controlled environments [7], [14],[22], [24] has con?rmed the feasibility of WiFi- -based vehicularInternet access for noninteractive applications. The possibilityand challenges to ... (QoS) assurance?can be achieved by acarefully planned sparse deployment of WiFi APs at roadside(see Fig. 1). In this paper, we study ... GO 769Fig. 1. Vehicular Internet access via roadside WiFi.providing roadside WiFi services. We envision a wireless ser-vice provider that implements a ... metric, called Contact Opportunity, as a charac-terization of a roadside WiFi network. Informally, the contactopportunity for a given deployment measures the ... two-stage stochastic optimizationapproach that minimizes the expected cost.While focusing on WiFi deployment, our study also providesuseful insights to the large deployment ... face of uncertainty. The de-ployment issues with respect to roadside WiFi networks havenot received much attention in the past. Our previous ...
... in various controlled environments [7], [14], [22],[24] and in situ WiFi networks [7], [10], [12], [22] have beenconducted, further con?rming the ... [10], [12], [22] have beenconducted, further con?rming the feasibility of WiFi- -basedvehicular Internet access for noninteractive applications. Inaddition to WiFi, , small cell architectures such as femtocells,which were initially designed ... of these efforts, scalable solutions for the deploymentand management of WiFi APs or femtocell base stations to en-able ef?cient vehicular Internet ...
... traf?c densityon road segment , i.e., the number of road-side WiFi serviceusers on per unit distance, denoted as , is within ...
... to each other are assigneddifferent channels to avoid interference (ns-3 WiFi does notmodel cross-channel interference). Each mobile node is con?g-ured with ...
... Columbus, OH,USA, and is free of potential interference from other WiFi net-works. The experiment was usually carried out at night whenthe ...
... Balasubramanian, R.Mahajan, A. Venkataramani, B. N. Levine, andJ. Zahorjan, ?Interactive WiFi connectivity for moving vehicles,? inProc. ACM SIGCOMM, Sep. 2008, pp. ... Eriksson, H. Balakrishnan, and S.Madden, ?Cabernet: vehicular con-tent delivery using WiFi, ,? in Proc. ACM MobiCom, Sep. 2008, pp.199?210.[13] I. Filippini, ... mining ap-proach,? in Proc. VLDB, Sep. 2007, pp. 794?805.[16] Google WiFi, , Mountain View, CA, USA, ?Google's Mountain ViewWiFi network,? [Online]. ...
... pp. 2831?2835.[38] Z. Zheng, P. Sinha, and S. Kumar, ?Sparse WiFi deployment for ve-hicular Internet access with bounded interconnection gap,? IEEE/ACMTrans. ...

Title:
WiFi positioning with propagation-based calibration

Keywords:
WiFi positioning

Abstract:
... as a cost-effective way to reduce the deployment cost of WiFi positioning systems. Contrary to traditional WiFi positioning systems, which require recording WiFi fingerprints together with ground truth locations, fingerprint generation attempts to ...

Full Text:
... as a cost-effective way to reduce the de-ployment cost of WiFi positioning systems. Contrary totraditional WiFi positioning systems, which require record-ing WiFi fingerprints together with ground truth locations,fingerprint generation attempts to automatically ... the dominant technology for in-door localization for several reasons: (i) WiFi enabled de-vices are commonplace and widely available; (ii) WiFi po-sitioning can take advantage of existing wireless infrastruc-tures without need ... without need for additional infrastructure investments;and (iii) the accuracy of WiFi positioning is sufficient formost practical indoor use cases. Amongst the ... formost practical indoor use cases. Amongst the myriad ofsolutions to WiFi positioning, fingerprinting has been themost successful due to its capability ... 14-16, 2015, Seattle, WA, USAACM 978-1-4503-3475-4/15/04.http://dx.doi.org/10.1145/2737095.2737144by recording large amounts of WiFi measurements togetherwith ground truth locations/labels. These measurementsare then used to ... reducing the calibration effort are essential for boostingthe uptake of WiFi positioning systems.Synthetic fingerprint generation has been proposed as acost-effective mechanism ...
... collated on a per-second basis, meaning we em-ulated receiving one WiFi fingerprint per second, which istraditionally the expected positioning rate. At ... rate weended up with 830 test fingerprints.We consider a probabilistic WiFi positioning system, orig-inally introduced by Roos et al. [4]. We ...

Keywords:
wifi roaming

Full Text:
... Aneasy way to achieve this goal could be to use WiFi networksalready deployed in most public and private places. How-ever, such ...

Title:
Vehicular wifi access and rate adaptation

Keywords:
wifi

Abstract:
<p>Vehicular WiFi access is distinct in two respects, (i) continuous mobility of ... develop a simple, low-overhead rate adaptation algorithm suited for vehicular WiFi

References:
J. Eriksson, H. Balakrishnan, and S. Madden. Cabernet: Vehicular Content Delivery Using WiFi. In ACM Mobicom, 2008.

Full Text:
speed_tpt.epsVehicular WiFi Access and Rate AdaptationAjinkya Uday Joshi and Purushottam KulkarniDepartment of ... and EngineeringIndian Institute of Technology Bombay, Mumbai, IndiaEmail: {ajinkyajo, puru}@cse.iitb.ac.inAbstract?Vehicular WiFi access is distinct in two re-spects, (i) continuous mobility of ... aimto develop a simple, low-overhead rate adaptation algorithmsuited for vehicular WiFi access.Categories and Subject Descriptors: C.2.1 [Computer-Communication Networks]:Network Architecture and Design[Wireless ... [Computer-Communication Networks]:Network Architecture and Design[Wireless communication]General Terms: Experimentation, Measurement, Algo-rithmsKeywords: WiFi, , Rate Adaptation, Mobility1. INTRODUCTIONUse of 802.11 (WiFi) )-based access for communication inoutdoor environments is on the rise?vehicle-to-vehicle ... [3] etc. In this paper, weconcern ourselves with the Vehicular WiFi Access scenario?use of roadside 802.11 installations (access points) to providecommunication ... due movement of people/objectsand interference due to other transmitting nodes.Vehicular WiFi access is distinct in two respects (i) con-tinuous mobility of ... 180 200 220 240RSSI(dBm)Distance (m)Figure 1: RSSI variation with vehicular WiFi access.and interference effects. [2] reports association durations of10 to 60 ... quality1There are several issues to be solved to instantiate Vehicu-lar WiFi
... AND FUTURE WORKPrevious work has identified and studied issues of WiFi- -based Vehicular access?association overhead, TCP through-put, TCP timeout durations and ...
... Eriksson, H. Balakrishnan, and S. Madden. Cabernet:Vehicular Content Delivery Using WiFi. . In ACM Mobicom,2008.[4] D. Hadaller, S. Keshav, T. Brecht, ...

Keywords:
WiFi

References:
DESHPANDE, P., HOU, X., AND DAS, S. R. Performance comparison of 3G and metro-scale WiFi for vehicular network access. In IMC, 2010.
MAHAJAN, R., ZAHORJAN, J., AND ZILL, B. Understanding WiFi-based Connectivity from Moving Vehicles. In Proc. ACM SIGCOMM, 2007.

Full Text:
... a wide-area wireless outdoortestbed [27]. We switched off the extra WiFi transmitters at thisecological reserve to obtain an interference-free network. In ...
... beaconing mes-sages as the underlying traffic to the fundamental characteristicsof WiFi- -based connectivity between BS and vehicles in urban set-tings. They ... done to develop and evalu-ate objective video quality metrics in WiFi networks, for examplein [5, 6], but their endeavors assume non-moving ...
... X., AND DAS, S. R. Performancecomparison of 3G and metro-scale WiFi for vehicularnetwork access. In IMC, 2010.[9] EFRON, B., AND TIBSHIRANI, ... 2009, pp. 92?105.[16] MAHAJAN, R., ZAHORJAN, J., AND ZILL, B.Understanding WiFi- -based Connectivity from MovingVehicles. In Proc. ACM SIGCOMM, 2007.[17] MATLOFF, ...

Title:
Video: WiFi-honk: smartphone-based beacon stuffed WiFi Car2X-communication system for vulnerable road user safety

Keywords:
wifi

Abstract:
... communication while walking. We propose a smartphone-based Car2X-communication system, named WiFi- -Honk, which can alert the potential collisions to both VRUs ... and vehicles in order to especially protect the distracted VRUs. WiFi- -Honk provides a practical safety means for the distracted VRUs ... for the distracted VRUs without requiring any special device using WiFi of smartphone. WiFi- -Honk removes the WiFi association overhead using the beacon stuffed WiFi communication with the geographic location, speed, and direction information of ... information of the smartphone replacing its SSID while operating in WiFi Direct/Hotspot mode, and also provides an efficient collision estimation algorithm ... issue appropriate warnings. Our experimental and simulation studies validate that WiFi- -Honk can successfully alert VRUs within a sufficient reaction time ...

References:
Kaustubh Dhondge, Sejun Song, Baek-Young Choi, Hyungbae Park, "WiFiHonk: Smartphone based Beacon Stuffed WiFi Car2X-Communication System for Vulnerable Road User Safety," IEEE 79th Vehicular Technology Conference (VTC), Spring 2014.

Full Text:
Microsoft Word - sysv08-dhondgeVideo: WiFi- -Honk ? Smartphone-based Beacon Stuffed WiFi Car2X-Communication System for Vulnerable Road User Safety Kaustubh Dhondge University ... communication while walking. We propose a smartphone-based Car2X-communication system, named WiFi- -Honk, which can alert the potential collisions to both VRUs ... and vehicles in order to especially protect the distracted VRUs. WiFi- -Honk provides a practical safety means for the distracted VRUs ... for the distracted VRUs without requiring any special device using WiFi of smartphone. WiFi- -Honk removes the WiFi association overhead using the beacon stuffed WiFi communication with the geographic location, speed, and direction information of ... information of the smartphone replacing its SSID while operating in WiFi Direct/Hotspot mode, and also provides an efficient collision estimation algorithm ... issue appropriate warnings. Our experimental and simulation studies validate that WiFi- -Honk can successfully alert VRUs within a sufficient reaction time ... and Subject Descriptors H.4.0 [Information Systems Applications]: General Keywords Smartphone; WiFi; ; Car2X-Communication REFERENCES [1] ?Honda Demonstrates Advanced Vehicle-to-Pedestrian and Vehicle-to-Motorcycle ... Song, Baek-Young Choi, Hyungbae Park, ?WiFiHonk: Smartphone based Beacon Stuffed WiFi Car2X-Communication System for Vulnerable Road User Safety,? IEEE 79th Vehicular ...