Energy and Resource Management Challenges in Data Centers

Tarek Abdelzaher, University of Illinois at Urbana Champaign


With the rise of "big data" analytics applications, a significant amount of load in cloud computing environments becomes data-centric. Computational workflows must be applied to real-time data streams in ways that respect throughput, timing, energy, and thermal constraints. This talk discusses underlying energy and resource management challenges in data centers, with an emphasis on those that arise from the data-intensive nature of modern computational workloads.  On the energy management side, solutions that work well for stateless servers (e.g., for Web servers serving short pages) no longer apply due to the stateful nature of data-intensive applications. Issues such as load prediction, data placement, memory cache management, and state consolidation become important problems if energy-efficient, energy-proportional operation is to be achieved. On the resource management side, meeting end-to-end time constraints becomes more complicated due to the non-independent nature of computational stages, linked together into complex workflows. Traditional deadline-scheduling becomes largely suboptimal at meeting end-to-end latency constraints. The talk describes recent solutions, related ongoing work, and experimental testing environments designed to address the above problems using a combination of optimization, prediction, and control techniques.

Presentation slides

Biography:  Tarek Abdelzaher received his B.Sc. and M.Sc. degrees in Electrical and Computer Engineering from Ain Shams University, Cairo, Egypt, in 1990 and 1994 respectively. He received his Ph.D. from the University of Michigan in 1999 on Quality of Service Adaptation in Real-Time Systems. He has been an Assistant Professor at the University of Virginia, where he founded the Software Predictability Group. He is currently a Professor and Willett Faculty Scholar at the Department of Computer Science, the University of Illinois at Urbana Champaign. He has authored/coauthored more than 170 refereed publications in real-time computing, distributed systems, sensor networks, and control. He is an Editor-in-Chief of the Journal of Real-Time Systems, and has served as Associate Editor of the IEEE Transactions on Mobile Computing, IEEE Transactions on Parallel and Distributed Systems, IEEE Embedded Systems Letters, the ACM Transaction on Sensor Networks, and the Ad Hoc Networks Journal. He chaired (as Program or General Chair) several conferences in his area including RTAS, RTSS, IPSN, Sensys, DCoSS, ICDCS, and ICAC. Abdelzaher's research interests lie broadly in understanding and controlling performance and temporal properties of networked embedded and software systems in the face of increasing complexity, distribution, and degree of embedding in an external physical environment. Tarek Abdelzaher is a recipient of the IEEE Outstanding Technical Achievement and Leadership Award in Real-time Systems (2012), the Xerox Award for Faculty Research (2011), as well as several best paper awards. He is a member of IEEE and ACM