Lightweight application-level crash consistency on transactional flash storage

摘要: Applications implement their own update protocols to ensure consistency of data on the file system. However, since current systems provide only a preliminary ordering guarantee, notably fsync, these become complex, slow, and error-prone. We present new system, CFS, that supports native interface for applications maintain crash data. Using can achieve by declaring code regions must operate atomically. By utilizing transactional flash storage (SSD/X-FTL), CFS lightweight mechanism consistency. Without using any heavyweight mechanisms based redundant writes ordering, atomically write multiple pages relevant metadata storage. We made three technical contributions develop with SSD/X-FTL in CFS: selective atomic propagation dirty pages, in-memory logging, delayed deallocation. Our evaluation five real-world shows CFS-based significantly outperform versions: 2-5× faster reducing disk 1.9-4.1× cache flushing 1.1-17.6×. Importantly, our porting effort is minimal: requires 317 lines modifications from 3.5 million ported applications.