MapperX: Adaptive Metadata Maintenance for Fast Crash Recovery of DM-Cache Based...
summary
Maps dirty-bit metadata into an on-disk adaptive bit-tree (ABT) that is synchronously updated per write but adapts its leaf granularity to exploit spatial locality, trading a tiny runtime cost for vastly faster recovery. Prototype is implemented in Linux DM-cache, reuses DM-cache metadata bits, and uses a 1s tuning period to add/delete ABT leaves. Evaluation with fio benchmark and public traces shows orders-of-magnitude crash-recovery speedup (baseline >2 hours) with negligible persistence overhead.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:14HDD-SSD Hybrid Storage
- 1:07DM-Cache
- 4:43Challenge
- 6:38MapperX Design
- 9:37Synchronous Bit-tree Update
- 10:17Periodical Adjust of ABT
- 12:06Fast Crash Recovery
- 12:41Implementation
- 13:44Micro Benchmarks
- 14:29Trace-Driven Evaluation
- 14:39Conclusion
related talks
13mBMC: Accelerating Memcached using Safe In-kernel Caching and Pre-stack Processing
14mArchTM: Architecture-Aware, High Performance Transaction for Persistent Memory
31mDesign and implementation of Autocaching for CXLSSD
19mP2CACHE: Exploring Tiered Memory for In-Kernel File Systems Caching
12mRethinking File Mapping for Persistent Memory
24mxadfs: Inverted DAX
48mPersistent Memory Usage within Linux Environment
21mAdvanced block storage test devices