CHESS: Systematic Stress Testing of Concurrent Software

Abstract

Concurrency is a fundamental attribute of systems software. Asynchronous computation is the norm in important software components such as operating systems, databases, and web servers. As multi-core architectures find their way into mainstream desktop computers, we are likely to see an increasing use of multithreading in application software as well. Unfortunately, the design of concurrent programs is a very challenging task. The main intellectual difficulty of this task lies in reasoning about the interaction between concurrently executing threads. Nondeterministic thread scheduling makes it extremely difficult to reproduce behavior from one run of the program to another. As a result, the process of debugging concurrent software becomes tedious resulting in a drastic decrease in the productivity of programmers. Since concurrency is both important and difficult to get right, it is imperative that we develop techniques and tools to automatically detect and pinpoint errors in concurrent programs.