Accelerating Nested Virtualization with HyperTurtle
summary
HyperTurtle reduces nested-VM world switches by offloading L1 code as eBPF running in L0, keeping L1 control while making only a few hundred non-intrusive kernel changes. It implements eBPF maps and helpers, a hook in L0, and per-exit eBPF handlers. Results show 5× faster EPT fault handling (EPT fault handling down to ~7 µs), up to 27% faster Kata container boot, ~72% memcached throughput gain, and near-native profiling latency.
AI-written summary · about
chapters (7)
- 0:00Nested virtualization and containers
- 1:43HyperTurtle acceleration approach
- 4:15Nested virtualization slowness
- 6:01HyperTurtle reduces world switch
- 7:51eBPF for virtual memory
- 11:28HyperTurtle performance gains
- 15:22Conclusion
related talks
37mEfficient and Flexible Virtual Machine Networking Through eBPF
11mUserspace Bypass: Accelerating Syscall-intensive Applications
31mKVM Nested Virtualization: Workings and Usecases
48mInto the hive of eBPF!
30mNested Virtualization (X86) Part II
30mNested Virtualization (X86) Part I
03h 01mHands-on Introduction to Virtualization
09mKVM Latency Performance Tuning