Description:

Adam Wierman

Caltech

Abstract: This talk will tell two parallel stories, one about designing sustainable data centers and one about the underlying algorithmic challenges, which fall into the context of online convex optimization. Story 1: The typical story surrounding data centers and energy is an extremely negative one: Data centers are energy hogs. This message is pervasive in both the popular press and academia, and it certainly rings true. However, the view of data centers as energy hogs is too simplistic. One goal of this talk is to highlight that, yes, data centers use a lot of energy, but data centers can also be a huge benefit in terms of integrating renewable energy into the grid and thus play a crucial role in improving the sustainability of our energy landscape. In particular, I will highlight a powerful alternative view: data centers as demand response opportunities. Story 2: Typically in online convex optimization it is enough to exhibit an algorithm with low (sub-linear) regret, which implies that the algorithm can match the performance of the best static solution in retrospect. However, what if one additionally wants to maintain performance that is nearly as good as the dynamic optimal, i.e., a good competitive ratio? In this talk, I'll highlight that it is impossible for an online algorithm to simultaneously achieve these goals. Luckily though, in practical settings (like data centers), noisy predictions about the future are often available, and I will show that, under a general model of prediction noise, even very limited predictions about the future are enough to overcome the impossibility result.
Biography: Adam Wierman is a Professor in the Department of Computing and Mathematical Sciences at the California Institute of Technology, where he is a founding member of the Rigorous Systems Research Group (RSRG) and maintains a popular blog called Rigor + Relevance. He received his Ph.D., M.Sc. and B.Sc. in Computer Science from Carnegie Mellon University in 2007, 2004, and 2001, respectively. His research interests center around resource allocation and scheduling decisions in computer systems and services. More specifically, his work focuses both on developing analytic techniques in stochastic modeling, queueing theory, scheduling theory, and game theory, and applying these techniques to application domains such as energy-efficient computing, data centers, social networks, and electricity markets.