Automatic data reuse for accelerating data intensive applications in the Cloud

摘要

This paper proposes a cost-oriented, petri-net based intelligent data reuse management to minimise data movement, reduce computing and storage costs and therefore improve the performance of data intensive applications in the Cloud. The proposed approach can automatically manage intermediate datasets of data intensive applications, predict the cost of the applications based on the previous runs, and intelligently decide whether the intermediate results should be stored or deleted for the subsequent runs, and dynamically optimise the cost of data intensive applications in the Cloud. The experimental evaluation demonstrates the proposed data reuse mechanism can significantly reduce the overall cost and enhance the performance of data intensive applications.