Tackling bpf uprobe performance

Session information has not yet been published for this event.

*

One Line Summary

Briefing the preliminary dyninst based study, and then discussions.

Abstract

Will present the initial dyninst based study to help improve performance of a subset of uprobe-attaching bpf programs. Pros and Cons will be presented. And then open the discussion on what are typical use cases, and whether dyninst approach is good or we need better one, etc.

Speaker

  • Biography

    I am a Facebook engineer and I am working on bcc/llvm/kernel support for BPF.