Accelerating Online Model Checking

Abstract

Online model checking is a lightweight verification technique to ensure at runtime the safety of the current execution trace of the system application under test. Doing model checking online suffers from the limited execution time allocated to each checking cycle. In this paper, we focus on accelerating online model checking so that as large the model space as possible can be explored in time. For this purpose, we introduce offline backward exploration so as to reduce the workload of online forward exploration. As a result, online model checking becomes online reach ability checking. SAT solver is used as verification engine for online model checking. We improve the performance of the SAT solver zChaff by optimizing and customizing zChaff with respect to the online model checking specific features. Moreover, we take advantage of the parallel feature and the multi-port memory available on FPGA chips. We present a new underlying architecture using 2 SAT solvers as verification engine for online model checking. We implement a quick prototype of the new underlying architecture for online model checking. Several experiments are done to test the performance of our online model checking.