A Novel approach for Dynamic Data Replication and Transfer in the Cloud to Reduce Redundancy and Cost

Abstract

As organisations chose cloud services over maintaining independent servers, there is an increased demand to reduce costs associated with such services. These services offer various storage tiers with different data access and storage prices. Manually selecting storage classes for each data object will be impractical and will result in non- optimal solutions. A system is needed to select and alter storage classes for data objects to ensure that objects reside in classes where residential costs and migration costs are minimal. This paper builds on previous work by employing a multi stage optimisation process. The first step is to use an optimal offline algorithm. As in previous work, this is followed by two online algorithms. Among online algorithms, the first algorithm is deterministic and the second is randomised. However, our proposal involves mapping the expected temporal changes of access rates of a data object. After access rates for an object drop below a certain point, they will be moved to a class with low storage costs and higher access costs. Moreover, this paper proposes storing data near its point of origin as a large part of digital content is media, which tends to be accessed near its point of origin.