Matrix-Stripe-Cache-Based Contiguity Transform for Fragmented Writes in RAID-5
IEEE Transactions on Computers
Efficient journaling writeback schemes for reliable and high-performance storage systems
Personal and Ubiquitous Computing
| Hi-index | 0.01 |
Small file accesses are still limited by disk head move-menton modern disk drives with the high disk bandwidth.Small file performance can be improved by grouping andclustering, each of which places multiple files in a direc-toryand places blocks of the same file on disks contigu-ously.These schemes make it possible for file systems touse large data transfers in accessing small files, reducingdisk accesses. However, as file systems become aged, disksbecome too fragmented to support the grouping and clus-teringof small files. This fragmentation makes it difficultfor file systems to take advantage of large data transfers,increasing disk I/Os. To offer a solution to this problem,we describe a De-fragmented File System (DFS). By usingdata cached in memory, DFS relocates and clusters datablocks of small fragmented files in a dynamic manner. Be-sides,DFS clusters related small files in the same directoryat contiguous disk locations. Measurements of DFS imple-mentationshow that the techniques alleviate file fragmenta-tionsignificantly and, in particular, performance for smallfile reads exceeds that of a traditional file system by 78%.