Object caching on Zoned Storage
summary
Proposes using Zoned Namespaces to lower write amplification for SSD caches by map large objects to zones and doing zone resets before reuse; experiments show sequential large-object placement matches ZNS performance but yields only modest gains. For small objects, random block writes hurt FTLs, so the proposal is to invalidate cold blocks instead of relocating them, accept cache misses, and reclaim zones round‑robin. Suggests prototypes as user‑space cache‑optimized FTLs or kernel knobs, and calls for realistic CacheLib/Cashflip workloads and benchmarks.
AI-written summary · about
chapters (10)
- 0:00Introduction to ZNS caching
- 1:07Caching challenges and solutions
- 3:34CashFlame caching system overview
- 5:00Large objects with ZNS performance
- 6:04Small objects and garbage collection
- 8:14Implementation approaches for ZNS caching
- 10:48Simple ZNS caching policy proposal
- 11:54Testing and benchmarking strategies
- 13:02Quality of service with ZNS
- 14:59Future directions and collaboration
related talks
31mFrom Open-Channel SSDs to Zoned Namespaces
14mZNS+: Advanced Zoned Namespace Interface for Supporting In-Storage Zone Compaction
14mUnleashing Zoned UFS: Cross-Layer Optimizations for Next-Generation Mobile Storage
20mZNSwap: un-Block your Swap
19mP2CACHE: Exploring Tiered Memory for In-Kernel File Systems Caching
12mRethinking File Mapping for Persistent Memory
25mShared CXL 3 memory what will be required
42mAllocators for Compressed Pages: In-depth Comparison with z3fold in Focus