Dynamic resource scheduling of cloud-based automatic test system using reinforcement learning
Yang Zhao, Xiao Ming-qing, Yawei Ge · 2017
Cloud-based auto-test is a promising technology in test field. In cloud-based auto-test, physical test resources are abstracted to logically manageable resource poll. The cloud-based automatic test system has to deal with multiple projects and resource utilizing requests simultaneously. Therefore, how to schedule the test resources dynamically and guarantee the long-term service quality is a key issue. This paper combines the idea of reinforcement learning with dynamic test resource scheduling to maximize the long-term performance of cloud-based automatic test system. A dynamic test resource scheduling model based on Markov decision process is built and a reinforcement learning algorithm based on Metropolis rule is proposed. According to the simulation results, the proposed algorithm shows good performance on 90 test instances. Moreover, the simulation results indicate that the proposed algorithm is more efficient in resource scheduling with more projects.