Dynamic Multi-Resource Optimization for Storage Acceleration in Cloud Storage Systems

Abstract

Demand for using cloud object storage has been increasing in order to efficiently manage a large number of binary large objects (BLOBs), including videos, photos and documents. Although many companies and institutions are currently trying to utilize public cloud object storage services such as AWS Simple Storage Service (S3), most of existing encoding systems for safe storage of data have not been optimized for current cloud object storage architecture. In this paper, we propose a novel dynamic extreme erasure encoding algorithm, namely DexEncoding aiming to maximize the utility of clients where the encoding locations in the cloud storage architecture are dynamically optimized between gateway and storage servers with respect to the time-varying cloud environment. Here, the utility captures the satisfaction of clients for the speed of data storage and fairness among clients. DexEncoding efficiently resolves resource bottlenecks by adapting to the dynamic network, processing and storage resource availability and storage request. Real measurement-driven simulations demonstrate that the proposed DexEncoding algorithm drastically outperforms that applied in the state-of-the-art object storage systems in a perspective of clients satisfaction. IEEE