Mining the potential opportunities for disk energy conservation

摘要

The energy conservation of disk drives depends on the idle intervals contained in I/O traffic. This paper proposes an algorithm for mining the two block correlations which contain long idle intervals and limited number of sporadic I/O requests. The mined idle intervals are long enough for energy conservation by switching the disk power states and filtering the sporadic I/O requests. Our experimental results with real-system traces show that the modern I/O traffics contain a lot of idle intervals which can be used to save energy.