Variation-aware resource allocation evaluation for cloud workflows using statistical model checking

Aiming at minimizing service operating costs and SLA (Service Level Agreement) violations, various resource allocation strategies have been investigated to support Cloud service providers' decision making. However, due to the service execution time variation, traditional optimal resource allocation strategies cannot achieve the best performance in practice. To address this problem, we propose an automated variation-aware evaluation framework for resource allocation strategies based on statistical model checker UPPAAL-SMC. Our framework can systematically evaluate the performance of resource allocation strategies under variations, and conduct complex queries on the quality of service. The experimental results show that our framework can not only filter inferior solutions efficiently, but also can enable the tuning of requirement constraints. Since our approach can be fully automated, the human efforts in resource allocation strategy evaluation can be significantly reduced.