Reproduce “eBeeMetrics: An eBPF-based Library Framework for Feedback-free Observability of QoS Metrics” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Muntaka Ibnath, Mohammadreza Rezvani, Daniel Wong
eBeeMetrics, an eBPF-based framework from the University of California, Riverside, offers accurate, real-time observability of application-level Quality-of-Service metrics, including latency and throughput, without needing application instrumentation. It achieved high fidelity tracking of client-reported metrics with minimal overhead, typically under a 3% latency increase, demonstrating its utility for system management runtimes.
Key insights
- Stable, protocol-specific identifiers, such as socket file descriptors or gRPC stream IDs, can reliably delineate individual request boundaries and enable per-request latency calculation from interleaved eBPF events.
- A hybrid probing strategy utilizing both kernel-level syscall probes and user-space application library probes is essential to transparently capture diverse client-server communication protocols, including HTTP/1.1 and multiplexed gRPC.
- Offloading complex data aggregation and statistical calculations to a user-space library, with minimal processing within eBPF kernel programs, maintains low runtime overhead and ensures the practical applicability of the observability framework.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.