Max: A Multicore-Accelerated File System for Flash Storage
summary
Max implements a multicore-friendly log-structured file system in the Linux kernel (based on F2FS) and demonstrates concrete speedups: evaluated on a 72-core machine with a high-bandwidth NVMe SSD, it shows up to 56× throughput for data overwrites, 18.6× improvement on metadata-heavy ops, and 2.9× vs. a recent multicore FS and 2.1× vs. F2FS. Max achieves this by adding a reader-writer semaphore that avoids cache-coherence traffic, a file cell layout to shard indexes and caches, and runtime-independent log partitions that defer ordering to recovery.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:22Background: Storage trend
- 0:58Background: Storage system
- 1:54Motivation: Multicore scalability
- 3:02Motivation: Analysis
- 4:38Max overview Key idea: sequential I/O (log structure) + efficient CPU use (sharding)
- 5:47LFS's concurrency control 1. Operations and concurrency control of classic LFS
- 7:14Reader Pass-through Semaphore
- 10:00File Cell
- 12:15Data and metadata scalability
- 12:58Application scalability
- 14:39Conclusion