Mining based file caching in hybrid system

In this work, we propose mining based file caching scheme for hybrid storage with NAND Flash and Hard disk drive. The objective of this research is to expedite the application launch latency in legacy Operating System. We apply data mining algorithm to extract the correlated file access pattern. In selecting the files to be loaded onto NAND flash, we incorporate the physical distance among the correlated file accesses so that the performance improvement can be maximized and apply greedy approach in selecting the files. When we incorporate the physical distance among the files on the disk, we can achieve upto 30% reduction on application lauch latency via relocating the selected files on the NAND Flash.