ACM SIGMETRICS 2021
June 14-18, 2021
The Unreasonable Effectiveness of the Lyapunov Drift Method: From Data Centers to Deep Reinforcement Learning
The Lyapunov drift method is a widely used tool to study the stability, convergence and performance of many dynamical systems. In this talk, we will present a survey of some recent techniques to identify the appropriate Lyapunov function to obtain the best performance bounds for two classes of problems: (i) resource allocation in data centers and (ii) reinforcement learning algorithms which use neural networks to approximate the value function.
R. Srikant is the Fredric G. and Elizabeth H. Nearing Professor in the Department of Electrical and Computer Engineering and the Coordinated Science Lab at the University of Illinois at Urbana-Champaign. He is also one of the two Co-Directors of the C3.ai Digital Transformation Institute, a consortium of universities including UIUC, Berkeley, MIT, UChicago, Princeton, Stanford, CMU and KTH and industries including C3.ai and Microsoft, whose goal is to promote research at the intersection of AI, ML and Cloud Computing. He is the recipient of the 2015 INFOCOM Achievement Award, the 2019 IEEE Koji Kobayashi Computers and Communications Award and the 2021 ACM SIGMETRICS Achievement Award. He has also received several Best Paper awards including the 2015 INFOCOM Best Paper Award, the 2015 WiOpt Best Paper Award and the 2017 Applied Probability Society Best Publication Award. He was the Editor-in-Chief of the IEEE/ACM Transactions on Networking from 2013-2017. His research interests include applied probability, stochastic control, machine learning and communication networks.
Stony Brook University
Sustainable IT and IT for Sustainability
Energy and sustainability have become one of the most critical issues facing our society. This talk presents my research efforts in making IT systems more sustainable, and furthermore, using IT to improve the sustainability of the energy system. In the first part, I will use Geographical Load Balancing as an example to illustrate how to exploit the spatial flexibility in cloud workloads for renewable energy integration and tackle the challenges of limited information and distributed control. Moving from theory to practice, I helped HP design and implement the industry’s first Net-Zero Energy Data Center and am currently working with IBM on their AI/ML systems. The second part focuses on data center demand response as an example. I will discuss its great potential and challenges, as well as my recent efforts in both control algorithm design for customers and market design for utility companies and our society. Future research directions will also be discussed.
Zhenhua Liu is an Assistant Professor of Operations Research in the Department of Applied Mathematics and Statistics and affiliated with the Department of Computer Science at Stony Brook University (SUNY at Stony Brook). He received his PhD in Computer Science from California Institute of Technology, under the supervision of Adam Wierman and Steven Low. His research aims to develop analytical models, theoretical results, and deployable algorithms to manage complex distributed systems with limited information and network constraints. He has helped HP design and implement the industry's first Net-zero Energy Data Center, which was named a 2013 Computerworld Honors Laureate. He was recently awarded an IBM 2020 Global University Program Academic Award for his research on resource management of AI/ML systems. His research work is widely cited and recognized in academia, including the Best Paper or Best Student Paper Awards at IEEE INFOCOM, ACM GREENMETRICS, and IEEE Green Computing Conference, the Pick of the Month award by IEEE STC on Sustainable Computing, a SPEC Distinguished Dissertation Award (honorable mention), an NSF CAREER award, and several Excellence in Teaching awards. For more information about Dr. Zhenhua Liu, please visit: http://www.ams.stonybrook.edu/~zhliu/.