Abstract
This article provides a detailed implementation study on the behavior of web serves that serve static requests where the load fluctuates over time (transient overload). Various external factors are considered, including WAN delays and losses and different client behavior models. We find that performance can be dramatically improved via a kernel-level modification to the web server to change the scheduling policy at the server from the standard FAIR (processor-sharing) scheduling to SRPT (shortest-remaining-processing-time) scheduling. We find that SRPT scheduling induces no penalties. In particular, throughput is not sacrificed and requests for long files experience only negligibly higher response times under SRPT than they did under the original FAIR scheduling.
- Abdelzaher, T. F. and Bhatti, N. T. 1999. Web content adaptation to improve server overload behavior. WWW8/Comput. Networks 31, 11-16, 1563--1577.]] Google Scholar
- Almesberger, W. Linux network traffic control---implementation overview. Available at http://lrcwww.epfl.ch/linux-diffserv/.]]Google Scholar
- Andresen, D. and Yang, T. 1997. Multiprocessor scheduling with client resources to improve the response time of WWW applications. In the International Conference on Supercomputing. 92--99.]] Google Scholar
- Arlitt, M. and Jin, T. 2000. Workload characterization of the 1998 world cup Web site. IEEE Network 14, 3, 30--37.]]Google Scholar
- Aron, M. and Druschel, P. 1999. TCP implementation enhancements for improving webserver performance. Tech. Rep. TR99-335, Rice University.]]Google Scholar
- Banga, G. and Druschel, P. 1999. Measuring the capacity of a web server under realistic loads. World Wide Web 2, 1-2, 69--83.]] Google Scholar
- Banga, G., Druschel, P., and Mogul, J. C. 1999. Resource containers: A new facility for resource management in server systems. In Proceedings of OSDI '99. 45--58.]] Google Scholar
- Bansal, N. and Harchol-Balter, M. 2001a. Analysis of SRPT scheduling: Investigating unfairness. In Proceedings of ACM SIGMETRICS '01.]] Google Scholar
- Bansal, N. and Harchol-Balter, M. 2001b. Scheduling solutions for coping with transient overload. Tech. rep. CMU-CS-01-134, Carnegie Mellon University.]]Google Scholar
- Barford, P. and Crovella, M. E. 1998. Generating representative Web workloads for network and server performance evaluation. In Proceedings of SIGMETRICS '98 (July) 151--160.]] Google Scholar
- Barford, P. and Crovella, M. E. 1999. A performance evaluation of hyper text transfer protocols. In Proceedings of ACM SIGMETRICS '99 (May) 188--179.]] Google Scholar
- Bender, M., Chakrabarti, S., and Muthukrishnan, S. 1998. Flow and stretch metrics for scheduling continuous job streams. In Proceedings of the 9th Annual ACM-SIAM Symposium on Discrete Algorithms.]] Google Scholar
- Bernstein, D. J. 1997. Syn cookies. http://cr.yp.to/syncookies.html.]]Google Scholar
- Bestavros, A., Carter, R. L., Crovella, M. E., Cunha, C. R., Heddaya, A., and Mirdad, S. A. 1995. Application-level document caching in the Internet. In Proceedings of the 2nd International Workshop on Services in Distributed and Networked Environments (SDNE'95) (June).]] Google Scholar
- Brakmo, L. and Peterson, L. 1995. Performance problems in 4.4 BSD TCP. ACM Comput. Comm. Rev. 25, 5.]] Google Scholar
- Braun, H. and Claffy, K. 1994. Web traffic characterization: An assessment of the impact of caching documents from NCSA's Web server. In Proceedings of the 2nd International WWW Conference.]]Google Scholar
- Cherkasova, L. and Phaal, P. 1998. Session-based admission control: A mechanism for improving the performance of an overloaded web server. Tech. Rep. HPL-98-119, Hewlett Packard Laboratories.]]Google Scholar
- Cockcroft, A. 1996. Watching your web server. The Unix Insider at http://www.unixinsider.com (April).]]Google Scholar
- Colajanni, M., Yu, P. S., and Dias, D. M. 1998. Analysis of task assignment policies in scalable distributed Web -server systems. IEEE Trans. Parall. Distrib. Syst. 9, 6, 585--699.]] Google Scholar
- Cooperative Association for Internet Data Analysis (CAIDA). 1999. Packet length distributions. http://www.caida.org/analysis/AIX/plen_hist.]]Google Scholar
- Crovella, M., Frangioso, R., and Harchol-Balter, M. 1999. Connection scheduling in web servers. In USENIX Symposium on Internet Technologies and Systems (Oct).]] Google Scholar
- Day, M., Cain, B., Tomlinson, G., and Rzewski, P. 2002. A model for content internetworking (cdi). Internet Draft (draft-ietf-cdi-model-02.txt) (May).]] Google Scholar
- Dias, D. M., Kish, W., Mukherjee, R., and Tewari, R. 1996. A scalable and highly available web server. In COMPCON. 85--92.]] Google Scholar
- Druschel, P. and Banga, G. 1996. Lazy receiver processing (LRP) : A network subsystem architecture for server systems. In Proceedings of OSDI '96 (Oct) 261--275.]] Google Scholar
- Elnikety, S., Nahum, E. M., Tracey, J., and Zwaenepoel, W. 2004. A method for transparent admission control and request scheduling in dynamic e-commerce Web sites. In International World-Wide Web Conference (WWW'04) (May) New York, NY.]] Google Scholar
- Feldmann, A. Web performance characteristics. IETF (Nov). http://www.research.att.com/anja/feldmann/papers.html.]]Google Scholar
- Friedman, E. J. and Henderson, S. G. 2003. Fairness and efficiency in web server protocols. In Proceedings of the ACM SIGMETRICS Conference on Measurement and Modeling of Computer Systems (May).]] Google Scholar
- Gong, M. and Williamson, C. 2003. Quantifying the properties of SRPT scheduling. In Proceedings of IEEE/ACM International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS).]]Google Scholar
- Gwertzman, J. and Seltzer, M. 1994. The case for geographical push-caching. In Proceedings of HotOS '94 (May).]] Google Scholar
- Harchol-Balter, M., Schroeder, B., Agrawal, M., and Bansal, N. 2003. Size-based scheduling to improve web performance. ACM Trans. Comput. Syst. 21, 2 (May).]] Google Scholar
- Internet Town Hall. The Internet traffic archives. Available at http://town.hall.org/Archives/pub/ITA/.]]Google Scholar
- Internet Traffic Report. 2004. http://www.internettrafficreport.com.]]Google Scholar
- IRCache Home. 2004. The trace files. http://www.ircache.net/Traces/.]]Google Scholar
- Iyer, R., Tewari, V., and Kant, K. 2000. Overload control mechanisms for web servers. In Workshop on Performance and QoS of Next Generation Networks (Nov).]]Google Scholar
- Johnson, K. L., Carr, J. F., Day, M. S., and Kaashoek, M. F. 2000. The measured performance of content distribution networks. In Proceedings of the 5th International Web Caching and Content Delivery Workshop.]]Google Scholar
- Kaashoek, M., Engler, D., Wallach, D., and Ganger, G. 1996. Server operating systems. In SIGOPS European Workshop '96. 141--148.]] Google Scholar
- Krishnamurthy, B. and Rexford, J. 2001. Web Protocols and Practice: HTTP/1.1, Networking Protocols, Caching, and Traffic Measurement. Addison-Wesley.]] Google Scholar
- Krishnamurthy, B., Wills, C., and Zhang, Y. 2001. The use and performance of content distribution networks. In Proceedings of the 1st ACM SIGCOMM Workshop on Internet Measurement.]] Google Scholar
- Kurose, J. and Ross, K. W. 2001. Computer Networking: A Top-Down Approach Featuring the Internet. Addison Wesley Longman, Inc.]] Google Scholar
- LeFebvre, W. 2002 Cnn.com: Facing a world crisis. The USENIX Technical Conference, (June).]]Google Scholar
- Maggs, B. Vice President of Research, 2001. Akamai Technologies. Personal communication.]]Google Scholar
- Manley, S. and Seltzer, M. 1997. Web facts and fantasy. In Proceedings of the 1997 USITS.]] Google Scholar
- McWherter, D., Schroeder, B., Ailamaki, A., and Harchol-Balter, M. 2004. Priority mechanisms for OLTP and transactional web applications. In the 20th International Conference on Data Engineering (ICDE'04).]] Google Scholar
- Microsoft TechNet Insights and Answer for IT Professional 2001. The arts and science of web server tuning with internet information services 5.0. http://www.microsoft.com/technet/.]]Google Scholar
- Mogul, J. C. 1995. The case for persistent-connection HTTP. In Proceedings of ACM SIGCOMM '95 (Oct). 299--313.]] Google Scholar
- Mogul, J. C. 1995. Network behavior of a busy Web server and its clients. Tech. Rep. 95/5, Digital Western Research Laboratory (Oct).]]Google Scholar
- Mogul, J. C. and Ramakrishnan, K. K. 1996. Eliminating receive livelock in an interrupt-driven kernel. In Proceedings of USENIX Technical Conference. 99--111.]] Google Scholar
- Murta, C. D. and Corlassoli, T. P. 2003. Fastest connection first: A new scheduling policy for web servers. In Proceedings of the 18th International Teletraffic Congress (ITC-18) (Sept).]]Google Scholar
- Nahum, E., Rosu, M., Seshan, S., and Almeida, J. 2001. The effects of wide-area conditions on www server performance. In Proceedings of ACM SIGMETRICS'01. 257--267.]] Google Scholar
- National Institute Standards and Technology. Nistnet. http://snad.ncsl.nist.gov/itg/nistnet/.]]Google Scholar
- Padhye, J., Firoiu, V., Towsley, D. F., and Kurose, J. F. 2000. Modeling tcp reno performance: A simple model and its empirical validation. IEEE/ACM Trans. Netw. 8, 2, 133--145.]] Google Scholar
- Pai, V. S., Druschel, P., and Zwaenepoel, W. 1999. Flash: An efficient and portable web server. In Proceedings of USENIX 1999 (June).]] Google Scholar
- Paxson, V. and Allman, M. 2000. Computing TCP's retransmission timer. RFC 2988, http://www.faqs.org/rfcs/rfc2988.html.]] Google Scholar
- Rai, I. A., Urvoy-Keller, G., and Biersack, E. 2003. Analysis of LAS scheduling for job size distributions with high variance. In Proceedings of the ACM Sigmetrics Conference on Measurement and Modeling of Computer Systems (SIGMETRICS) (June).]] Google Scholar
- Rawat, M. and Kshemkayani, A. 2003. SWIFT: Scheduling in web servers for fast response time. In the 2nd IEEE International Symposium on Network Computing and Applications (April).]] Google Scholar
- Rizzo, L. 1997. Dummynet: A simple approach to the evaluation of network protocols. ACM Comput. Commun. Rev. 27, 1.]] Google Scholar
- Schrage, L. E. and Miller, L. W. 1966. The queue M/G/1 with the shortest remaining processing time discipline. Operations Resear. 14, 670--684.]]Google Scholar
- Seshan, S., Balakrishnan, H., Padmanabhan, V. N., Stemm, M., and Katz, R. 1998. TCP behavior of a busy internet server: Analysis and improvements. In Proceedings of the Conference on Computer Communications (IEEE Infocom). 252--262.]]Google Scholar
- Silberschatz, A., Galvin, P., and Gagne, G. 2002. Operating System Concepts, 6th Ed. John Wiley & Sons.]] Google Scholar
- Stallings, W. 2001. Operating Systems, 4th Ed. Prentice Hall.]]Google Scholar
- The World Wide Web Consortium (W3C). Libwww---the W3C protocol library. http://www.w3.org.]]Google Scholar
- Voigt, T. and Gunnigberg, P. 2001. Kernel-based control of persistent web server connections. ACM SIGMETRICS Performance Evaluation Review 29, 2, 20--25.]] Google Scholar
- Voigt, T., Tewari, R., Freimuth, D., and Mehra, A. 2001. Kernel mechanisms for service differentiation in overloaded web servers. In Proceedings of the USENIX Annual Technical Conference (June) Boston, MA.]] Google Scholar
- Welsh, M. and Culler, D. 2003. Adaptive overload control for busy internet servers. In Proceedings of the USENIX Symposium on Internet Technologies and Systems.]] Google Scholar
- Wierman, A. and Harchol-Balter, M. 2003. Classifying scheduling policies with respect to unfairness in an M/GI/1. In ACM Sigmetrics International Conference on Measurement and Modeling of Computer Systems.]] Google Scholar
Index Terms
Web servers under overload: How scheduling can help
Recommendations
Size-based scheduling to improve web performance
Is it possible to reduce the expected response time of every request at a web server, simply by changing the order in which we schedule the requests? That is the question we ask in this paper.This paper proposes a method for improving the performance of ...
Controlling overload in networks of SIP servers
ICNP '08: Proceedings of the 2008 IEEE International Conference on Network ProtocolsThe session initiation protocol (SIP) is rapidly being adopted as the signaling protocol for establishing, modifying and terminating multimedia sessions. With the increasing use of SIP in large deployments, it is now becoming apparent that the current ...
Classifying scheduling policies with respect to unfairness in an M/GI/1
It is common to evaluate scheduling policies based on their mean response times. Another important, but sometimes opposing, performance metric is a scheduling policy's fairness. For example, a policy that biases towards small job sizes so as to minimize ...








Comments