Project · Multi-Endpoint Latency Profiler

project

hard

module project

Ship something real. Submit your work when you're done.

Brief

Write a CLI script that accepts a list of API endpoints (via a text file or command-line arguments) and, for each one, runs 100 sequential requests, records per-request latency, and outputs a formatted table showing p50, p95, p99, max, and request count per endpoint. If the response includes a Server-Timing header, parse and display each named phase alongside the total. The script must accept a --threshold flag (default 500ms); any endpoint whose p99 exceeds the threshold is highlighted as a warning.

Deliverables

A single runnable script in any language from the tabs (Go, Python, Rust, or Node.js) that reads a list of URLs and benchmarks each one.
Output table showing endpoint, n, p50, p95, p99, max, and a WARN flag for endpoints exceeding the threshold.
Server-Timing header parsing: if the response includes it, a second table row shows each named phase (e.g. db=42ms, auth=3ms).
A README with: how to run, what the columns mean, and one observed result from running it against a real (or public) API.

How we grade it

Benchmark results are repeatable: running the script twice against the same endpoint produces p50 values within 20% of each other.
Server-Timing parsing correctly handles both `dur=` values and bare tokens (e.g. `cache;desc=miss` with no duration).
The --threshold flag correctly flags any endpoint whose p99 exceeds it — tested by passing an artificially low value (e.g. 1ms).
The script handles connection errors gracefully (a failed request is counted as a timeout, not a crash).

Project · Multi-Endpoint Latency Profiler

Hints

Expected output

Stretch goals