Brief Announcement: Distributed Synthesis of Fault-Tolerance

摘要

Synthesis algorithms usually suffer from two factors of time and space complexity. In order to overcome the time complexity problem, several approaches have been proposed in the literature to incrementally add properties to existing verified programs (e.g., [1]). In order to overcome the space explosion problem, recently, an increasing interest in parallel and distributed techniques has emerged in the model checking community (e.g., [2,3]). Such techniques parallelize the state space of a given model over a network or cluster of workstations and run a distributed model checking algorithm over the parallelized state space. On the other hand, the space explosion problem is still unaddressed in the context of automated program synthesis. With this motivation, we concentrate on the problem of designing distributed algorithms for automated program synthesis. More specifically, we parallelize two synthesis algorithms (from [1]) for adding two levels of fault-tolerance, namely failsafe and masking, to existing fault-intolerant programs. We assume that programs are in the high atomicity model, where all processes can read and write all the program variables in one atomic step.