Abstract
Many modern computing systems must provide reliable latency with minimal energy. Two central challenges arise when allocating system resources to meet these conflicting goals: (1) complexity modern hardware exposes diverse resources with complicated interactions and (2) dynamics latency must be maintained despite unpredictable changes in operating environment or input. Machine learning accurately models the latency of complex, interacting resources, but does not address system dynamics; control theory adjusts to dynamic changes, but struggles with complex resource interaction. We therefore propose CALOREE, a resource manager that learns key control parameters to meet latency requirements with minimal energy in complex, dynamic en- vironments. CALOREE breaks resource allocation into two sub-tasks: learning how interacting resources affect speedup, and controlling speedup to meet latency requirements with minimal energy. CALOREE deines a general control system whose parameters are customized by a learning framework while maintaining control-theoretic formal guarantees that the latency goal will be met. We test CALOREE's ability to deliver reliable latency on heterogeneous ARM big.LITTLE architectures in both single and multi-application scenarios. Compared to the best prior learning and control solutions, CALOREE reduces deadline misses by 60% and energy consumption by 13%.
- Jason Ansel, Maciej Pacula, Yee Lok Wong, Cy Chan, Marek Olszewski, Una-May O'Reilly, and Saman Amarasinghe. 2012. Siblingrivalry: online autotuning through local competitions. In CASES. Google Scholar
Digital Library
- Jason Ansel, Yee Lok Wong, Cy Chan, Marek Olszewski, Alan Edelman, and Saman Amarasinghe. 2011. Language and compiler support for auto-tuning variable-accuracy algorithms. In CGO. Google Scholar
Digital Library
- R. M. Bell, Y. Koren, and C. Volinsky. 2008. The BellKor 2008 solution to the Netlix Prize. Technical Report. ATandT Labs.Google Scholar
- C. Bienia, S. Kumar, J. P. Singh, and K. Li. 2008. The PARSEC Benchmark Suite: Characterization and Architectural Implications. In PACT. Google Scholar
Digital Library
- Ramazan Bitirgen, Engin Ipek, and Jose F. Martinez. 2008. Coordinated management of multiple interacting resources in chip multiprocessors: A machine learning approach. In MICRO. Google Scholar
Digital Library
- Giorgio C Buttazzo, Giuseppe Lipari, Luca Abeni, and Marco Caccamo. 2006. Soft Real-Time Systems: Predictability vs. Efficiency: Predictability vs. Efficiency. Springer. Google Scholar
Digital Library
- Shuai Che, Michael Boyer, Jiayuan Meng, David Tarjan, Jeremy W. Sheafer, Sang-Ha Lee, and Kevin Skadron. 2009. Rodinia: A Benchmark Suite for Heterogeneous Computing. In IISWC. Google Scholar
Digital Library
- Jian Chen and Lizy Kurian John. 2011. Predictive coordination of multiple on-chip resources for chip multiprocessors. In ICS. Google Scholar
Digital Library
- Jian Chen, Lizy Kurian John, and Dimitris Kaseridis. 2011. Modeling Program Resource Demand Using Inherent Program Characteristics. SIGMETRICS Perform. Eval. Rev. 39, 1 (June 2011), 1-12. Google Scholar
Digital Library
- Ryan Cochran, Can Hankendi, Ayse K. Coskun, and Sherief Reda. 2011. Pack&Cap: adaptive DVFS and thread packing under power caps. In MICRO. Google Scholar
Digital Library
- Sarah Dean, Horia Mania, Nikolai Matni, Benjamin Recht, and Stephen Tu. 2017. On the Sample Complexity of the Linear Quadratic Regulator. Technical Report 1710.01688v1. arXiv.Google Scholar
- Christina Delimitrou and Christos Kozyrakis. 2013. Paragon: QoS-aware Scheduling for Heterogeneous Datacenters. In ASPLOS. Google Scholar
Digital Library
- Christina Delimitrou and Christos Kozyrakis. 2014. Quasar: Resource-efficient and QoS-aware Cluster Management. In ASPLOS. Google Scholar
Digital Library
- Zhaoxia Deng, Lunkai Zhang, Nikita Mishra, Henry Hoffmann, and Fred Chong. 2017. Memory Cocktail Therapy: A General Learning-Based Framework to Optimize Dynamic Tradeoffs in NVM. In MICRO. Google Scholar
Digital Library
- Christophe Dubach, Timothy M. Jones, Edwin V. Bonilla, and Michael F. P. O'Boyle. 2010. A Predictive Model for Dynamic Microarchitectural Adaptivity Control. In MICRO. Google Scholar
Digital Library
- Antonio Filieri, Henry Hoffmann, and Martina Maggio. 2014. Automated design of self-adaptive software with control-theoretical formal guarantees. In ICSE. Google Scholar
Digital Library
- Antonio Filieri, Henry Hofmann, and Martina Maggio. 2015. Automated multi-objective control for self-adaptive software design. In FSE. Google Scholar
Digital Library
- J. Flinn and M. Satyanarayanan. 1999. Energy-aware adaptation for mobile applications. In SOSP. Google Scholar
Digital Library
- Jason Flinn and M. Satyanarayanan. 2004. Managing battery lifetime with energy-aware adaptation. ACM Trans. Comp. Syst. 22, 2 (May 2004). Google Scholar
Digital Library
- Rodrigo Fonseca, Prabal Dutta, Philip Levis, and Ion Stoica. 2008. Quanto: Tracking Energy in Networked Embedded Systems. In OSDI. Google Scholar
Digital Library
- Andrew Gelman, John B Carlin, Hal S Stern, David B Dunson, Aki Vehtari, and Donald B Rubin. 2013. Bayesian data analysis. CRC press.Google Scholar
- Ashvin Goel, David Steere, Calton Pu, and Jonathan Walpole. 1998. SWiFT: A Feedback Control and Dynamic Reconiguration Toolkit. In 2nd USENIX Windows NT Symposium. Google Scholar
Digital Library
- Matthew Halpern, Yuhao Zhu, and Vijay Janapa Reddi. {n. d.}. Mobile CPU's rise to power: Quantifying the impact of generational mobile CPU design trends on performance, energy, and user satisfaction. In HPCA.Google Scholar
- Joseph L. Hellerstein, Yixin Diao, Sujay Parekh, and Dawn M. Tilbury. 2004. Feedback Control of Computing Systems. John Wiley&Sons. Google Scholar
Digital Library
- Henry Hofmann. 2015. JouleGuard: energy guarantees for approximate applications. In SOSP. Google Scholar
Digital Library
- Henry Hofmann, Anant Agarwal, and Srinivas Devadas. 2012. Selecting Spatiotemporal Patterns for Development of Parallel Applications. IEEE Trans. Parallel Distrib. Syst. 23, 10 (2012). Google Scholar
Digital Library
- Henry Hofmann, Jonathan Eastep, Marco D. Santambrogio, Jason E. Miller, and Anant Agarwal. 2010. Application Heartbeats: a generic interface for specifying program performance and goals in autonomous computing environments. In ICAC. Google Scholar
Digital Library
- Henry Hofmann, Jim Holt, George Kurian, Eric Lau, Martina Maggio, Jason E. Miller, Sabrina M. Neuman, Mahmut Sinangil, Yildiz Sinangil, Anant Agarwal, Anantha P. Chandrakasan, and Srinivas Devadas. 2012. Self-aware computing in the Angstrom processor. In DAC. Google Scholar
Digital Library
- T. Horvath, T. Abdelzaher, K. Skadron, and Xue Liu. 2007. Dynamic Voltage Scaling in Multitier Web Servers with End-to-End Delay Control. Computers, IEEE Transactions on 56, 4 (2007). Google Scholar
Digital Library
- Connor Imes, David H. K. Kim, Martina Maggio, and Henry Hoffmann. 2015. POET: A Portable Approach to Minimizing Energy Under Soft Real-time Constraints. In RTAS.Google Scholar
- Engin Ipek, Onur Mutlu, José F. Martínez, and Rich Caruana. 2008. Self-Optimizing Memory Controllers: A Reinforcement Learning Approach. In ISCA. Google Scholar
Digital Library
- Syed Muhammad Zeeshan Iqbal, Yuchen Liang, and Hakan Grahn. 2010. ParMiBench - An Open-Source Benchmark for Embedded Multi-processor Systems. IEEE Comput. Archit. Lett. 9, 2 (July 2010). Google Scholar
Digital Library
- C. Karamanolis, M. Karlsson, and X. Zhu. 2005. Designing controllable computer systems. In HotOS. Berkeley, CA, USA. Google Scholar
Digital Library
- David H. K. Kim, Connor Imes, and Henry Hoffmann. 2015. Racing and Pacing to Idle: Theoretical and Empirical Analysis of Energy Optimization Heuristics. In CPSNA. Google Scholar
Digital Library
- Minyoung Kim, Mark-Oliver Stehr, Carolyn Talcott, Nikil Dutt, and Nalini Venkatasubramanian. 2013. xTune: A Formal Methodology for Cross-layer Tuning of Mobile Embedded Systems. ACM Trans. Embed. Comput. Syst. 11, 4 (Jan. 2013). Google Scholar
Digital Library
- Etienne Le Sueur and Gernot Heiser. 2011. Slow Down or Sleep, That is the Question. In Proceedings of the 2011 USENIX Annual Technical Conference. Portland, OR, USA. Google Scholar
Digital Library
- B.C. Lee, J. Collins, Hong Wang, and D. Brooks. 2008. CPR: Composable performance regression for scalable multiprocessor models. In MICRO. Google Scholar
Digital Library
- Benjamin C. Lee and David Brooks. 2008. Eiciency Trends and Limits from Comprehensive Microarchitectural Adaptivity. In ASPLOS. Google Scholar
Digital Library
- Benjamin C. Lee and David M. Brooks. 2006. Accurate and Efficient Regression Modeling for Microarchitectural Performance and Power Prediction. In ASPLOS. Google Scholar
Digital Library
- Matthew Lentz, James Litton, and Bobby Bhattacharjee. 2015. Drowsy Power Management. In SOSP. Google Scholar
Digital Library
- W.S. Levine. 2005. The control handbook. CRC Press.Google Scholar
- Baochun Li and K. Nahrstedt. 1999. A control-based middleware framework for quality-of-service adaptations. IEEE Journal on Selected Areas in Communications 17, 9 (1999). Google Scholar
Digital Library
- J. Li and J.F. Martinez. 2006. Dynamic power-performance adaptation of parallel computation on chip multiprocessors. In HPCA.Google Scholar
- Lennart Ljung. 1999. System Identiication: Theory for the User. Prentice Hall PTR, Upper Saddle River, NJ, USA. Google Scholar
Digital Library
- C. Lu, Y. Lu, T.F. Abdelzaher, J.A. Stankovic, and S.H. Son. 2006. Feedback Control Architecture and Design Methodology for Service Delay Guarantees in Web Servers. IEEE TPDS 17, 9 (September 2006), 1014--1027. Google Scholar
Digital Library
- Martina Maggio, Henry Hofmann, Alessandro V. Papadopoulos, Jacopo Panerati, Marco D. Santambrogio, Anant Agarwal, and Alberto Leva. 2012. Comparison of Decision-Making Strategies for Self-Optimization in Autonomic Computing Systems. ACM Trans. Auton. Adapt. Syst. 7, 4, Article 36 (Dec. 2012), 32 pages. Google Scholar
Digital Library
- M. Maggio, H. Hofmann, M. D. Santambrogio, A. Agarwal, and A. Leva. 2013. Power Optimization in Embedded Systems via Feedback Control of Resource Allocation. IEEE Transactions on Control Systems Technology 21, 1 (Jan 2013).Google Scholar
Cross Ref
- Martina Maggio, Alessandro Vittorio Papadopoulos, Antonio Filieri, and Henry Hofmann. 2017. Automated Control of Multiple Software Goals Using Multiple Actuators. In ESEC/FSE. Google Scholar
Digital Library
- John D. McCalpin. 1995. Memory Bandwidth and Machine Balance in Current High Performance Computers. IEEE TCCA Newsletter (Dec. 1995), 19--25.Google Scholar
- Nikita Mishra. 2017. Statistical Methods for Improving Dynamic Scheduling and Resource Usage in Computing Systems. Ph.D. Dissertation. https://search.proquest.com/docview/1928485902?accountid=14657Google Scholar
- Nikita Mishra, Connor Imes, Huazhe Zhang, John D Lafferty, and Henry Hoffmann. 2016. Big Data for LITTLE Cores: Combining Learning and Control for Mobile Energy Efficiency. Technical Report TR-2016-10. University of Chicago, Dept. of Comp. Sci.Google Scholar
- Nikita Mishra, Huazhe Zhang, John D. Laferty, and Henry Hoffmann. 2015. A Probabilistic Graphical Model-based Approach for Minimizing Energy Under Performance Constraints. In ASPLOS. Google Scholar
Digital Library
- Akihiko Miyoshi, Charles Lefurgy, Eric Van Hensbergen, Ram Rajamony, and Raj Rajkumar. 2002. Critical Power Slope: Understanding the Runtime Effects of Frequency Scaling. In ICS. Google Scholar
Digital Library
- Carl N Morris. 1983. Parametric empirical Bayes inference: theory and applications. J. Amer. Statist. Assoc. 78, 381 (1983), 47-55.Google Scholar
Cross Ref
- Adam J. Oliner, Anand P. Iyer, Ion Stoica, Eemil Lagerspetz, and Sasu Tarkoma. 2013. Carat: Collaborative Energy Diagnosis for Mobile Devices. In Proceedings of the 11th ACM Conference on Embedded Networked Sensor Systems (SenSys '13). ACM, New York, NY, USA, Article 10, 14 pages. Google Scholar
Digital Library
- Sinno Jialin Pan and Qiang Yang. 2010. A Survey on Transfer Learning. IEEE Trans. on Knowl. and Data Eng. 22, 10 (Oct. 2010), 1345--1359. Google Scholar
Digital Library
- Paula Petrica, Adam M. Izraelevitz, David H. Albonesi, and Christine A. Shoemaker. 2013. Flicker: A Dynamically Adaptive Architecture for Power Limited Multicore Systems. In ISCA. Google Scholar
Digital Library
- Dmitry Ponomarev, Gurhan Kucuk, and Kanad Ghose. 2001. Reducing Power Requirements of Instruction Scheduling Through Dynamic Allocation of Multiple Datapath Resources. In MICRO. Google Scholar
Digital Library
- Raghavendra Pothukuchi, Amin Ansari, Petros Voulgaris, and Josep Torrellas. 2016. Using Multiple Input, Multiple Output Formal Control to Maximize Resource Efficiency in Architectures. In ISCA. Google Scholar
Digital Library
- Ramya Raghavendra, Parthasarathy Ranganathan, Vanish Talwar, Zhikui Wang, and Xiaoyun Zhu. 2008. No "power" struggles: coordinated multi-level power management for the data center. In ASPLOS. Google Scholar
Digital Library
- R. Rajkumar, C. Lee, J. Lehoczky, and Dan Siewiorek. 1997. A resource allocation model for QoS management. In RTSS. Google Scholar
Digital Library
- Arjun Roy, Stephen M. Rumble, Ryan Stutsman, Philip Levis, David Mazières, and Nickolai Zeldovich. 2011. Energy Management in Mobile Devices with the Cinder Operating System. In EuroSys. Google Scholar
Digital Library
- Muhammad Husni Santriaji and Henry Hofmann. 2016. GRAPE: Minimizing energy for GPU applications with performance requirements. In MICRO.Google Scholar
- Akbar Sharifi, Shekhar Srikantaiah, Asit K. Mishra, Mahmut Kandemir, and Chita R. Das. 2011. METE: meeting end-to-end QoS in multicores through system-wide resource management. In SIGMETRICS. Google Scholar
Digital Library
- Kai Shen, Arrvindh Shriraman, Sandhya Dwarkadas, Xiao Zhang, and Zhuan Chen. 2013. Power Containers: An OS Facility for Fine-grained Power and Energy Management on Multicore Servers. SIGPLAN Not. 48, 4 (March 2013), 65-76. Google Scholar
Digital Library
- David C. Snowdon, Etienne Le Sueur, Stefan M. Petters, and Gernot Heiser. 2009. Koala: A Platform for OS-level Power Management. In EuroSys. Google Scholar
Digital Library
- Michal Sojka, Pavel Písa, Dario Faggioli, Tommaso Cucinotta, Fabio Checconi, Zdenek Hanzálek, and Giuseppe Lipari. 2011. Modular software architecture for lexible reservation mechanisms on hetero- geneous resources. Journal of Systems Architecture 57, 4 (2011). Google Scholar
Digital Library
- Srinath Sridharan, Gagan Gupta, and Gurindar S. Sohi. 2013. Holistic Run-time Parallelism Management for Time and Energy Efficiency. In ICS. Google Scholar
Digital Library
- David C. Steere, Ashvin Goel, Joshua Gruenberg, Dylan McNamee, Calton Pu, and Jonathan Walpole. 1999. A Feedback-driven Proportion Allocator for Real-rate Scheduling. In Proceedings of the Third Symposium on Operating Systems Design and Implementation (OSDI '99). USENIX Association, Berkeley, CA, USA, 145--158. http://dl.acm.org/citation.cfm?id=296806.296820 Google Scholar
Digital Library
- Q. Sun, G. Dai, and W. Pan. 2008. LPV Model and Its Application in Web Server Performance Control. In ICCSSE. Google Scholar
Digital Library
- G. Tesauro. 2007. Reinforcement Learning in Autonomic Computing: A Manifesto and Case Studies. IEEE Internet Computing 11 (2007). Issue 1. Google Scholar
Digital Library
- Michel Tokic. 2010. Adaptive ∈-Greedy Exploration in Reinforcement Learning Based on Value Diferences. In KI. Google Scholar
Digital Library
- Stephen Tu and Benjamin Recht. 2017. Least-Squares Temporal Difference Learning for the Linear Quadratic Regulator. Technical Report 1712.08642v1. arXiv.Google Scholar
- Vibhore Vardhan, Wanghong Yuan, Albert F. Harris III, Sarita V. Adve, Robin Kravets, Klara Nahrstedt, Daniel Grobe Sachs, and Douglas L. Jones. 2009. GRACE-2: integrating fine-grained application adaptation with global adaptation for saving energy. IJES 4, 2 (2009).Google Scholar
- Greg Welch and Gary Bishop. {n. d.}. An Introduction to the Kalman Filter. Technical Report TR 95-041. UNC Chapel Hill, Department of Computer Science. Google Scholar
Digital Library
- Jonathan A. Winter, David H. Albonesi, and Christine A. Shoemaker. 2010. Scalable thread scheduling and global power management for heterogeneous many-core architectures. In PACT. Google Scholar
Digital Library
- Qiang Wu, Philo Juang, Margaret Martonosi, and Douglas W. Clark. 2004. Formal online methods for voltage/frequency control in multiple clock domain microprocessors. In ASPLOS. Google Scholar
Digital Library
- Weidan Wu and Benjamin C Lee. 2012. Inferred models for dynamic and sparse hardware-software spaces. In Microarchitecture (MICRO), 2012 45th Annual IEEE/ACM International Symposium on. IEEE, 413--424. Google Scholar
Digital Library
- Joshua J. Yi, David J. Lilja, and Douglas M. Hawkins. 2003. A Statistically Rigorous Approach for Improving Simulation Methodology. In HPCA. Google Scholar
Digital Library
- Wanghong Yuan and Klara Nahrstedt. 2003. Energy-efficient soft real-time CPU scheduling for mobile multimedia systems. In SOSP. Google Scholar
Digital Library
- Huazhe Zhang and Henry Hofmann. 2016. Maximizing Performance Under a Power Cap: A Comparison of Hardware, Software, and Hybrid Techniques. In ASPLOS. Google Scholar
Digital Library
- R. Zhang, C. Lu, T.F. Abdelzaher, and J.A. Stankovic. 2002. Control-Ware: A middleware architecture for Feedback Control of Software Performance. In ICDCS. Google Scholar
Digital Library
- Xiao Zhang, Rongrong Zhong, Sandhya Dwarkadas, and Kai Shen. 2012. A Flexible Framework for Throttling-Enabled Multicore Management (TEMM). In ICPP. Google Scholar
Digital Library
- Yanqi Zhou, Henry Hofmann, and David Wentzlaf. 2016. CASH: Supporting IaaS Customers with a Sub-core Conigurable Architecture. In ISCA. Google Scholar
Digital Library
- Yuhao Zhu and Vijay Janapa Reddi. 2013. High-performance and energy-efficient mobile web browsing on big/little systems. In HPCA. Google Scholar
Digital Library
Index Terms
CALOREE: Learning Control for Predictable Latency and Low Energy
Recommendations
Generative and multi-phase learning for computer systems optimization
ISCA '19: Proceedings of the 46th International Symposium on Computer ArchitectureMachine learning and artificial intelligence are invaluable for computer systems optimization: as computer systems expose more resources for management, ML/AI is necessary for modeling these resources' complex interactions. The standard way to ...
CALOREE: Learning Control for Predictable Latency and Low Energy
ASPLOS '18: Proceedings of the Twenty-Third International Conference on Architectural Support for Programming Languages and Operating SystemsMany modern computing systems must provide reliable latency with minimal energy. Two central challenges arise when allocating system resources to meet these conflicting goals: (1) complexity modern hardware exposes diverse resources with complicated ...
Minimizing energy under performance constraints on embedded platforms: resource allocation heuristics for homogeneous and single-ISA heterogeneous multi-cores
Special Issue on the 4th Embedded Operating Systems Workshop (EWiLi 2014)This paper explores the problem of energy optimization in embedded platforms. Specifically, it studies resource allocation strategies for meeting performance constraints with minimal energy consumption. We present a comparison of solutions for both ...







Comments