RaceTest: harmful data race detection based on testing technology in WS-BPEL

Abstract

Web service composition is widely used nowadays, and it is realized by Web Service-Business Process Execution Language (WS-BPEL). WS-BPEL provides concurrency mechanisms so that data races commonly occur within it. Some researchers have focus on data race detection in WS-BPEL. However, none of them focus on identifying harmful data races from benign data races. This paper proposed RaceTest, a novel technique that aims to identify harmful data races in WS-BPEL effectively. Unlike previous detectors, RaceTest identifies harmful data races through dynamic detection based on testing technology. Firstly, it obtains potential data races with static detection. Secondly, it prunes the false positives statically with constraint solving and then generates test cases that contain feasible potential races. Finally, it executes test cases to identify harmful data races by deterministic technique which can actively control thread interleaving. The experimental results show that RaceTest is effective; it effectively reduces a lot of false positives and confirms harmful data races with high accuracy.