HTTP Request Benchmark

Server assisted Performance

A responsible micro-benchmark: up to 50 requests with concurrency up to 5, timeouts and rate limits, reporting requests/second, latency distribution and error rate. Not a load-testing or stress service.

Privacy: This tool sends the request through the Mutqan backend relay to reach the target. Only the target URL, headers and body you enter are transmitted, and nothing is stored.

How to use HTTP Request Benchmark

  1. Enter a public endpoint URL and choose the HTTP method.
  2. Set the number of requests (≤ 50) and concurrency (≤ 5); add headers or a body if needed.
  3. Click "Run benchmark" and wait a few seconds for the relay to finish.
  4. Read throughput and percentiles; check the error rate and status breakdown before trusting the numbers.
  5. For a real load test, generate k6/JMeter scripts with the Concurrent Request Test Generator.

HTTP Request Benchmark features

  • Bounded micro-benchmark: up to 50 requests with concurrency 1–5
  • Requests per second, error rate and full latency percentiles (p50–p99)
  • Custom method, headers and JSON body for authenticated or write endpoints
  • Latency histogram and percentile chart for quick visual comparison
  • Status breakdown (200 ×48, 429 ×2) so rate limiting is obvious
  • Raw samples downloadable as CSV

HTTP Request Benchmark example

20 requests, concurrency 2

Input:

GET https://api.example.com/products
Requests: 20 · Concurrency: 2

Output:

Requests: 20 · Requests/s: 9.4 · Error rate: 0%
Average: 205 ms · p50: 198 ms · p90: 262 ms · p95: 281 ms · p99: 310 ms
Min / Max: 160 ms / 310 ms · Consistency: 71/100

Frequently asked questions about HTTP Request Benchmark

What are the limits and why?

At most 50 requests and concurrency 5 per run, with request timeouts and per-IP rate limits. This keeps the tool safe for third-party endpoints — it is a micro-benchmark, not a load-testing service.

How is requests/second calculated?

The relay measures wall-clock throughput for the whole run; when that figure is unavailable the tool derives it from the samples and the concurrency level.

Can I send a body and headers?

Yes. Add headers one per line and a body for POST/PUT/PATCH. Never paste production credentials — use test tokens.

What does the error rate include?

Failed connections, timeouts and non-2xx statuses. The status breakdown (e.g. 200 ×48, 429 ×2) shows whether rate limiting or errors skewed the numbers.

How do I run a real load test?

Use the Concurrent Request Test Generator to produce k6, JMeter, Locust, Artillery or Node.js scripts and run them from your own infrastructure against systems you are authorised to test.

Technical notes

The relay executes the run with a fixed concurrency and returns raw samples plus its own throughput figure, which is measured over wall-clock time and therefore accounts for concurrency correctly. Local statistics are recomputed from the samples so the same percentile definition is used across all performance tools.

The limits (50 requests, concurrency 5, per-IP rate limits and timeouts) are enforced server-side. They make the tool safe to point at third-party APIs but also mean it cannot reveal capacity limits — that is the job of a proper load test run from your own infrastructure.