A multi-policy adaptive scheduling framework in virtual clouds

With the development of cloud computing technology, many cloudbased data centres are beginning to connect with each other aiming at building large-scale federated cloud platforms. In such a federated cloud platform, the efficiency and effectiveness of traditional schedulers will be significantly degraded due to the unpredictable resource availability and high networking latency. In this paper, we propose a multi-policy adaptive scheduling framework, which is capable of evaluating the scheduling scheme made by existing schedulers. In this way, the proposed scheduling framework can make full use of the advantages of existing scheduling policies and avoid their shortcomings, and therefore makes the final scheduling decisions more efficient and effectiveness. In addition, the proposed scheduling framework is designed as extensible component that can be easily extended by incorporating other scheduling policies. The prototype of this scheduling framework is tested in a real-world federated cloud platform. The experimental results show that it can significantly improve the quality-of-service for the underlying cloud infrastructures as well as cloud user's satisfactory.