A partial replication load balancing algorithm for distributed Data as a Service (DaaS)

This paper presents an easy and direct algorithm to solve the issue of load balancing in providing Data as a Service (DaaS) in the Cloud. The algorithm is based on some earlier approach for efficient dual direction data downloading. Our contribution is in solving the issue of the high storage demand when storing replicated data in multiple Cloud nodes. Rather than storing full replicas, our approach devises a model to store partial replicas of the data on multiple distributed Cloud servers. Moreover, we provide a direct method to download this data from multiple Cloud servers by coordinating the download process among the different nodes and different parts of the replicas on the Cloud. We implemented and evaluated our algorithm on the top of the CloudSim toolkit.