The dominant paradigm in resource allocation for cloud computing is that customers arrive, request resources, and relinquish them some unknown amount of time later.  This creates challenging resource allocation decisions for cloud providers given the lack of understanding of customer plans.  These decisions can be made more efficiently if cloud providers have additional information from customers.  Economics provides a set of tools for understanding how to elicit information from customers while understanding that they have an incentive to be strategic in the way they approach the system.  I’ll give examples from several lines of work within this space including reserving capacity for fixed lengths of time, caching for serverless computing, and fair multi-resource allocation.

Bio:

Ian Kash is an associate professor in the computer science department at the University of Illinois Chicago and visiting the Argonne Leadership Computing Facility during Fall 2024 for a sabbatical.  Previously he was a researcher at Microsoft Research and a postdoc at the Center for Research on Computation and Society at Harvard.  He completed his Ph.D. In computer science from Cornell.

See upcoming and previous presentations at CS Seminar Series.