Understanding Ruby with BPF

Transcript

1. rbperf: Understanding Ruby with BPF Javier Honduvilla Coto <javierhonduco@fb.com> October 28th-29th, 2020
2. Why BPF?
3. Why BPF? - Flexibility
4. Why BPF? - Flexibility - Low overhead
5. Why BPF? - Flexibility - Low overhead - Continuous profiling
6. Why BPF? - Flexibility - Low overhead - Continuous profiling - No modifications of the tracee
7. rbperf
8. rbperf - Profile Ruby programs
9. rbperf - Profile Ruby programs - Trace complex Ruby programs execution
10. rbperf – on-CPU profiling - $ rbperf record --pid=124 cpu - $ rbperf report [...]
11. rbperf – Rails on-CPU profile
12. rbperf – tracing write(2) calls - $ rbperf record \ --pid=124 event \ --tracepoint=syscalls:sys_enter_write - $ rbperf report [...]
13. Architecture 2. Event (timer, syscall, etc) BPF code (bpf/rbperf.c) Read frame Driver (rbperf.py) 1. Adds info (pid to profile, thread address) 3. Receives stacktrace 4. Serialisation and persistence BPF tail-calls Bounded loop
14. Challenges - Implementing the stack walking for a dynamic language
15. Challenges - Implementing the stack walking for a dynamic language - Supporting multiple Ruby versions
16. Challenges - Implementing the stack walking for a dynamic language - Supporting multiple Ruby versions - Correctness testing
17. Challenges - Implementing the stack walking for a dynamic language - Supporting multiple Ruby versions - Correctness testing - BPF safety features
18. Future plans - Integrate in Facebook’s profiling infra - Rewrite OSS driver program - Make the OSS version awesome - Better documentation (including how to measure overhead) - Add more output formats - Open source GDB / drgn helper - Other tools? - Containers support? - Support request-oriented workloads?
19. Thanks! :) https://github.com/facebookexperimental/rbperf javierhonduco@fb.com @javierhonduco