It has long been known that in order to minimize mean delay, one should prioritize small jobs – specifically one should use Shortest Remaining Processing Time first (SRPT). However, looking around systems today, SRPT scheduling is almost never used. Why? The answer is not so mysterious: Designers worry that favoring small jobs could lead to starvation of large jobs. Designers don't have access to exact size information. The list goes on. In this talk we will examine the validity of many of these worries.
Adam Wierman is an Assistant Professor of Computer Science at the California Institute of Technology, where he is a member of the Lee Center for Advanced Networking, the Center for the Mathematics of Information, and the Social and Information Sciences Laboratory. He received his PhD (2007) and MS (2004) in Computer Science from Carnegie Mellon University under the supervision of Mor Harchol-Balter. He also briefly served as a visiting researcher at the EURANDOM institute under the supervision of Onno Boxma and Ivo Adan.
Wierman's research interests are in improving computer system design through the use of measurement and modeling. His main focus is on the scheduling and resource allocation decisions in distributed systems; however, he has also been involved with the design of manufacturing systems and telecommunication protocols.
Catherine Copetas, copetas [atsymbol] cs.cmu.edu