GA Based Placement Optimization for Hybrid Distributed Storage

The VOD (video on-demand) applications, in order to minimize access latency and improve quality of service, require periodically optimizing the placement of their large volume of video, courseware, and database files in data centers. Migration-enhanced data placement schemes allow data movement at runtime, which imposes significant overhead, particularly for hybrid distributed storage systems that consist of both SSDs and HDDs. In this paper, we propose a genetic algorithm based multi-objective optimization scheme for data placement and migration optimization in software defined storage systems. We prototype and evaluate the proposed scheme with comparison to simple heuristic designs. Our results show that our design acquires better nondominated resolution set and achieves around 15% access delay and migration cost reduction.