TOSS: Traffic-aware distributed object-based storage using software-defined networks

Abstract

As storage systems grow to Petascale, the demand for object storage increases. In a large scale heterogeneous object storage systems, efficient selection of storage targets for placing objects is critically important since the performance depends on even distribution of objects across storage targets. An efficient storage target selection for object placement depends not only on available storage target capacity but also network bandwidth. The storage target selection based either only on the available storage capacity or only on the available network bandwidth may not result in the optimal usage of storage/network resources to achieve the performance. The object placement under heterogeneous environment considering load balancing is a challenging problem. There is a need to orchestrate the network and storage resources with efficient object to storage mapping. In this paper, we present an efficient and scalable object placement strategy using software-defined networking (SDN) technique. We demonstrate the effectiveness of our method through simulation results and compare with Distributed Hash Table (DHT) method.