NyxCache: Flexible and Efficient Multi-tenant Persistent Memory Caching
summary
NyxCache is an access-regulation framework for multi-tenant persistent-memory (PM) caches that implements runtime microcontrolled experiments to identify which cache causes interference, plus lightweight request regulation and per-cache PM-usage accounting. It is a pure-software, black-box design deployed on PMM devices and claims up to 5x better performance isolation than bandwidth-limiting and up to 6x higher throughput for best-effort caches versus prior DRAM-based approaches. Implementations show write-spike isolation, accurate per-cache PM usage limits, proportional sharing and ≈2x improved fairness in slowdown.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 0:40Memorybased today
- 1:37Inmemory keyword caches
- 2:38PMM throughput
- 3:13Cloud resource limiting approach
- 5:07Design new mechanism
- 9:31DRM interface analysis
- 12:56Demonstration
- 15:02Summary
related talks
17mPreview: Persistent Memory
17mScalaCache: Scalable User-Space Page Cache Management with Software-Hardware...
36mCaching Strategies for an Erlang Based Web Stack
53mLeveraging Caching For Cloud Native Applications
35mDeveloping Software for Persistent Memory
49mCache à la carte: a framework for in-memory caching
17mPushing the Limits of In-Network Caching for Key-Value Stores
43mPromise and Pitfalls of Persistent Memory