GCASR 2016‎ > ‎Presentations‎ > ‎

Learning from the Past: Fast, On-Demand Analysis of Prior Executions with Eidetic Systems

Learning from the Past: Fast, On-Demand Analysis of Prior Executions with Eidetic Systems

Jason Flinn (University of Michigan, Ann Arbor)



Abstract:

Eidetic computer systems can recall any past state and replicate the execution that produced that state.  Eidetic systems enable powerful queries about past executions for security forensics, understanding data provenance, debugging, and troubleshooting applications.  I will discuss how we can provide eidetic support in commodity operating systems at reasonable storage and performance overheads, and how we can utilize computer clusters to parallelize complex, introspective queries over program executions so as to enable interactive analysis of past executions.

Bio:

Jason Flinn is a Professor of Computer Science and Engineering at the University of Michigan, Ann Arbor.  His research interests include operating systems, distributed systems, and mobile computing.  He is currently the director of the Software Systems Laboratory at the University of Michigan.  His research has been recognized with an NSF CAREER award and 8 best paper awards at SOSP, OSDI, ASPLOS, FAST, and MobiSys.
Comments