Decentralized, Epoch-based F2FS Journaling with Fine-grained Crash Recovery
summary
Claims a working prototype that journals only metadata, embeds logs into inodes, and uses an epoch-based control/data decoupling so new transactions do not wait for ongoing I/O. Implementation modifies F2FS with per-inode log lists, a 256 MB log area, and a fast-forward-to-latest apply that consolidates small updates. Evaluated with fio and Android-like workloads, it reduces checkpoint time up to 4.9× and average operation latency up to 35%, while lowering lock contention and journal storage overhead.
AI-written summary · about