ReplayDebugging

From RAD Lab

Revision as of 17:10, 14 January 2006; view current revision
←Older revision | Newer revision→
Jump to: navigation, search

Contents

Replay Debugging for Distributed Applications

Students

  • Dennis Geels
  • Gautam Altekar

Summary

We have developed a new debugging tool for deployed distributed applications, liblog. Our goal is to help diagnose and repair software errors that are difficult to detect before deployment, including complex race conditions and so-called "heisenbugs". To this end we have extended deterministic replay technology to support continuous logging of C/C++ applications running on hundreds of machines. Technological contributions include user-level logging of multi-threaded applications, consistent replay of UDP and TCP traffic among arbitrary network subgroups, and integration with GDB for source-level debugging of multiple processes.

Project Website

http://research.geels.org:8080/