I/O in a Flash: Evolution of ONTAP to Low-Latency SSDs
summary
Claims that ONTAP’s read path is refactored to hit tens of microseconds with SSDs by eliminating message-passing hops and context switches. Implements MP-safe page headers, an IO buffer type, a storage location cache and hierarchical attributes cache so protocol threads can look up mappings and dispatch DMA straight to storage threads. Results show much lower latency and higher throughput for 4K random and 64K sequential reads while retaining ONTAP’s read-after and atomicity guarantees.
AI-written summary · about
chapters (7)
- 0:00Introduction and context
- 2:11Initial system foundations
- 3:52Read path analysis
- 7:45Optimizing the read path
- 13:54Advanced caching strategies
- 17:39Correctness and consistency
- 19:27Conclusion and takeaways
related talks
19mAsynchronous I/O Stack: A Low-latency Kernel I/O Stack for Ultra-Low Latency SSDs
19mBest I/O is No I/O
17mλ-IO: A Unified IO Stack for Computational Storage
21mStreamCache: Revisiting Page Cache for File Scanning on Fast Storage Devices
38mScaling IO-bound application in the cloud
04mArtur on SSD's
25mRecent developments in I/O control and testing
40mAchieving the Ultimate Performance with KVM
