Optimization of mass storage systems by quality of service

Abstract

With the explosive growth of digital information, storage systems are becoming larger and more complicated. Especially in large scale storage systems, they must be able to deliver satisfactory I/O performance (availability, reliability, etc.) under both expected and unexpected workloads, and handle data with real timeliness requirements. However, typical management methods of mass storage systems cannot meet these requirements. One solution is using Quality of Service (QoS) approaches in object-based storage systems. These QoS approaches include Quality of Storage Service (QoSS) framework and QoSS improvements. After researching the QoS requirements and the various data access characterizations in different applications, we designed a QoSS management framework. And then according to this framework, we gave some improvements to efficiently realize QoSS in object-based storage systems. To demonstrate the effectiveness of our proposed solution, we developed a prototype of Attribute-managed Storage System with QoSS (AM-QoSS) by extending both OSD and iSCSI protocols. The measurement results finally show that the proposed QoS management approach achieves optimization of storage systems, such as higher performance and better guaranteed QoS.