CAREER: Language Features for Robust Software

University Of California-Irvine, Irvine CA

Investigators

Abstract

Developing reliable software systems is a key challenge for our society. Despite many years of effort, software systems still suffer from catastrophic failures. All too often, software failures are caused by the propagation of errors through critical components of the system. Unfortunately, current software development tools and practices actually encourage the introduction of unnecessary dependencies that serve to propagate errors between conceptually unrelated components. The goal of this research is to explore a new approach for creating software systems that are robust against failures. The approach exposes the high-level structure of a software system to the compiler and run-time environment to enable the automatic application of techniques that appropriately manage error propagation to make software systems more resilient. The foundation of this approach is a specification language that allows the developer to identify tasks and describe when tasks should be invoked and how each task changes the conceptual states of objects. A set of techniques use this information to monitor and analyze task execution to eliminate unnecessary dependencies. The broader impact of this research is the potential to mitigate the effects of software faults.