This 16-minute conference talk from USENIX FAST '25 introduces ScaleLFS, a log-structured file system designed with scalable garbage collection for commodity SSDs. Learn about three innovative techniques that enhance sustained performance: per-core dedicated garbage collectors that parallelize operations, a scalable victim manager for concurrent metadata updates, and a scalable victim protector enabling page-level garbage collection procedures. Presented by researchers from Seoul National University and Chung-Ang University, the talk demonstrates how ScaleLFS outperforms existing systems like F2FS by up to 3.5×, other scalable LFS implementations by 4.6×, and parallel GC schemes by 7.0×. Discover how these improvements can significantly boost sustained performance on commodity solid-state drives.
Overview
Syllabus
FAST '25 - ScaleLFS: A Log-Structured File System with Scalable Garbage Collection for Commodity///
Taught by
USENIX