A novel virtual machine scheduling policy based on performance prediction model

In cloud platforms, virtual machine scheduling policy plays an important role for providing desirable service quality for users. In many existing scheduling policies, the task execution time is often assumed to be constant or defined by users. However, either unpredictable workload or resource unreliability may significantly affect task execution time, which in turn results in inefficient scheduling decisions. In this paper, we first present a task execution time model by applying queue theory; then we use this model to predict the performance of application at runtime and propose a novel virtual machine scheduling policy. By conducting extensive experiments, we investigate the effectiveness and efficiency of the proposed scheduling policy. The experimental results indicate that it can significantly reduce the response time of cloud application comparing with other existing scheduling policies.