The Next Chapter in the Sordid Love/Hate Relationship Btwn DBs & OSes
summary
Claims that eBPF user-bypass reduces OS overhead are backed by concrete builds and numbers: a BPF-based proxy called tiger rewrites hot packet-forwarding paths and yields ~40% higher throughput on the smallest VM, forcing an Odyssey deployment to use an eight-times larger machine to match it. BPF-DB implements an in-kernel, ACID, multi-versioned key-value store with up to ~8 versions (verifier limits), uses 32 tail calls and a circular buffer + user-space flusher for durability, and shows lower P99 latency than competitors.
AI-written summary · about
chapters (8)
- 0:00Introduction to DB/OS conflict
- 3:15Why databases hate operating systems
- 7:26Kernel bypass vs user bypass
- 10:14Building proxies with BPF
- 17:00BPF DB architecture
- 26:30BPF DB performance results
- 29:40Future of BPF in databases
- 31:05Q&A
related talks
32meBPF: BPF kernel infrastructure
14mWill eBPF Save Us From the Next Global Outage?
21mPerformance Wins with BPF: Getting Started
43mHotBPF++: A More Powerful Memory Protection for the Linux Kernel
48mInto the hive of eBPF!
16mElectrode: Accelerating Distributed Protocols with eBPF
36mThe journey of BPF from restricted C language towards extended and safe C.
40mPanel Discussion: eBPF: A New Era in Cloud Infrastructure Tools