Improving Linux Performance Debugging with Nusku: A Rust and eBPF Profiler

Why does performance debugging still suck so much?

Developer Ali Amer is building Nusku, an open-source performance profiler developed using Rust and eBPF. The tool aims to replicate the intuitive Visual Studio debugging experience on Linux systems by providing instant, live CPU and memory metrics.

Why This Matters

While Linux offers powerful tools like perf and bpftrace, they often suffer from steep learning curves and complex configurations that hinder immediate troubleshooting. Nusku addresses this gap by providing a low-overhead, ‘run-and-see’ profiler that works on production systems to visualize performance spikes exactly when they occur, bridging the gap between expert-level tools and intuitive developer experiences.

Key Insights

• Nusku leverages eBPF perf events to sample CPU usage and capture user-space stack traces with minimal system overhead.
• The tool utilizes Blazesym for symbolization, converting raw addresses into function names and source file lines in real-time.
• Aggregation logic in Rust allows for immediate identification of ‘hot’ functions within a live terminal view.
• The project roadmap includes integrating the Ratatui library for an advanced terminal user interface and native flamegraph generation.
• Nusku is designed for zero-config operation, supporting direct PID attachment or running commands like ‘nusku ./my-app’.

Working Examples

Planned CLI interface for Nusku profiler

nusku --pid 1234
nusku ./my-app
nusku -c node app.js

Current real-time terminal output of the Nusku profiler

── PID 132612 ── 98 samples ── CPU 99.0% ── RSS 2.0 MiB ── VIRT 3.1 MiB ──
% COUNT FUNCTION SOURCE ADDRESS
────────────────────────────────────────────────────────────────────────────────────────────────
27.6% 27 <core::ops::range::Range<T> as … range.rs:773 0x00005efa8cfd4b95
11.2% 11 testing::hot_c main.rs:7 0x00005efa8cfd4cc7
10.2% 10 <i32 as core::iter::range::Step… range.rs:197 0x00005efa8cfd4acd

Practical Applications

• Production Profiling: Attaching Nusku to a live PID to identify function-level CPU spikes without service restarts. Pitfall: Using heavy profilers in production can cause latency; Nusku uses eBPF to minimize this impact.
• Local Development: Running a binary directly through Nusku to verify memory allocations and CPU jumps during specific UI interactions. Pitfall: Relying on static analysis alone misses dynamic performance bottlenecks identified by live sampling.

References:

• https://dev.to/aliamerj/why-does-performance-debugging-still-suck-so-much-17j2
• https://github.com/aliamerj/nusku