VeriLin: A Linearizability Checker for Large-Scale Concurrent Objects

Qiaowen Jia, Yi Lv, Peng Wu, Bohua Zhan, Jifeng Hao, Hong Ye, Chao Wang

Abstract: Linearizability is an important correctness criterion for concurrent objects, and there have been several existing tools for checking linearizability. However, due to the inherent exponential complexity of the problem, existing tools have difficulty scaling up to large, industrial-sized concurrent objects. In this paper, we introduce VeriLin, a new linearizability checker that incorporates a more general checking algorithm as well as associated testing strategies, that allow it to continue to be effective for large-scale concurrent objects and long histories. For evaluation, we apply VeriLin to checking linearizability of student implementations of a train ticketing system, as well as the task management and scheduling module of a proprietary multicore operating system.