Skip to main content

Automated Program Repair for Distributed Systems

Publication Type
Year of Publication
Conference/Journal Name
ACM SIGACT News Distributed Computing Column
Model repair is a formal method that aims at fixing bugs in models automatically. Typically, these models are finite state automata that can be compactly represented using guarded commands or variations thereof. The bugs in these models can be identified using traditional techniques, such as verification, testing, or runtime monitoring. However, these techniques do not assist in fixing bugs
automatically. The goal in model repair is to automatically transform an input model into another model that satisfies additional properties (e.g., a property that the original model fails to satisfy). Moreover, such transformation should preserve the existing specification of the input model. In this article, we review the efforts in the past decade on developing model repair algorithms in different domains. These domains include distributed computing, fault-tolerance and self-stabilization, and real-time systems. We present the results on complexity analysis, techniques for tackling intractability of the problem and scalability, and related tools. The techniques and tools discussed in this article demonstrate the feasibility of automated synthesis of well-known protocols such as Byzantine agreement, token ring, fault-tolerant mutual exclusion, etc.