A distributed architecture for intra- and inter-cloud data management

Abstract

When envisioning "The Cloud," one is often presented with an idyllic black box of functionality that seamlessly stores arbitrary amounts of data while providing endless CPU cycles. However, Cloud deployments are often partitioned along institutional and middleware boundaries that link computation and storage infrastructures. Decoupling storage from computation resource providers allows for greater flexibility in resource provisioning and new data storage paradigms to emerge. This paper proposes a decentralized data management architecture that facilitates interoperability between Clouds and other heterogeneous systems. The goals of this research are to augment the latent storage capacity within provisioned Cloud VMs with existing institutional resources to build low-cost Storage Clouds for scientific computing.