Description: Thousands of server farm “load balancing” policies balance the load among servers, but is load balancing a good thing? Why is the donor machine given all the control in cycle stealing? Can favoring “short” jobs in scheduling policies such as Shortest-Remaining-Processing-Time-First (SRPT) hurt “long” ones? Computer systems design is based on many commonly held beliefs and heuristics, but many those beliefs have never been challenged. Mor Harchol-Balter, associate professor of computer science at Carnegie Mellon University, explores some fundamental questions in system design in an attempt to demonstrate how new research in analytical performance modeling helps us overturn some age-old beliefs.
Speaker(s):
Mor Harchol-Balter, associate professor, Computer Science, Carnegie Mellon University
|