Feasibility, Efficiency, and Effectiveness of Self-Optimizing Storage Systems
Recent work has proposed making intelligent use of data access patterns for building self-optimizing storage systems. However, despite the continued increase in the CPU-I/O performance gap, such systems are far from wide adoption. The authors argue that the key reason for the lack of real systems adopting this novel idea is that current studies leave several key questions unanswered. The pinpoint the research and practical challenges that must be addressed in building effective self-optimizing storage systems. The answers to these questions are based on practical experience while building such a system.