Skip to content

load-test-baseline.md

Latest commit

 

History

History
44 lines (37 loc) · 1.5 KB

load-test-baseline.md

File metadata and controls

44 lines (37 loc) · 1.5 KB
Raw

Load-test baseline

This is a small local baseline for the Docker Compose API. It is intended to show how the measurement works, not to claim production capacity.

Headline result: all 20 requests succeeded at 4.8279 requests/s, with a 0.0% error rate, 669.3392 ms p50 latency, and 1368.4134 ms p95 latency.

Measurement Result
Endpoint POST /predict
Input assets/dataset/non_defect_sample.jpg
Requests 20
Concurrency 4
Successful requests 20
Failed requests 0
Error rate 0.0%
Throughput 4.8279 requests/s
Mean latency 803.3074 ms
p50 latency 669.3392 ms
p95 latency 1368.4134 ms

The test ran on 2026-08-31 against http://127.0.0.1:8000/predict, with the API and PostgreSQL running through Docker Compose. Latency is measured from the client before the multipart request is sent until the response body is read, so it includes upload, API processing, database persistence, and the response transfer.

The reusable command is:

.venv\Scripts\python.exe scripts\load_test.py `
  --url http://127.0.0.1:8000/predict `
  --image assets\dataset\non_defect_sample.jpg `
  --requests 20 `
  --concurrency 4 `
  --output artifacts\load-test\report.json

The full machine-readable result is written to artifacts/load-test/report.json, which is intentionally ignored because it is generated output. To compare configurations, change --requests or --concurrency and record the new environment and values alongside the result.