Frontend-only arena demo to benchmark and compare reliability of HTTP clients under identical chaos conditions:
- native
fetch axioskyffetch
The app wraps requests with chaos rules (latency, random failures, rate limiting) and runs a repeatable workload against a test endpoint, measuring reliability scores, latency, and error patterns side-by-side.
- Install dependencies
npm install - Start dev server
npm run dev - Open
index.htmlin your browser (served at http://localhost:5173 by default)
- Responsive UI with preset scenarios and real-time progress
- Chaos controls: latency range, random failures, rate limiting, throttling
- Client configuration: independent timeout, retry, and plugin settings per client
- Run comparison: side-by-side reliability score, error distribution, and latency metrics
- Exportable results: download full run snapshots as JSON (schema v2)
- Reliability Score: weighted metric accounting for success rate, timeouts, and tail latency
- Error Distribution: breakdown by status code and error type
- Latency Percentiles: p50 and p95 response times
- Throughput: requests per second
- Attempt Tracking: total transport attempts (including retries)
- Requested concurrency is a target, but browsers apply their own connection pooling limits (~6 per origin)
- Chaos is applied at the transport layer, before client libraries handle retries
- Each run is fully isolated—transport stats are per-client and don't carry over