Load Test Report 2024‐01‐18
This load test is performed to verify the fix in S3Save of uptime-service-backend. Fix provided in:
- PR MinaProtocol: https://github.com/MinaProtocol/mina/pull/14918
- PR Uptime-Service-Backend: https://github.com/MinaFoundation/uptime-service-backend/pull/35
Specs: Linux (PopOS 22.04), 64GB RAM, CPU: 13th Gen Intel i9-13900HX (32) @ 5.200GHz
Toolset: Locust tool, payload with dummy block and snark work. Payload is fairly big = 14MB. The test was performed using following script: https://github.com/MinaFoundation/mina-delegation-program-tech/tree/08f23252cf62c79017ff2cb345223df143815784/load_test
Test Conditions: 100 vusers, 1s rampup, duration: 10m, payload ~14MB.
Notes: For the sake of test sections of // wl := h.app.Whitelist.ReadWhitelist() and // payload, err := req.Data.MakeSignPayload() have been commented out in the code, as the test payload consists of dummy block and snark work. Test payload is fairly big (14MB) which probably exceeds normal payload significantly, however the idea is to put system under load.
Locust report: https://minafoundation.github.io/mina-delegation-program-tech/load_test/results/2024-01-18/1.0-pre-fix/
Locust report: https://minafoundation.github.io/mina-delegation-program-tech/load_test/results/2024-01-18/1.0-post-fix/
It appears that fix significantly improves both response times (See Locust reports RPS (0.4 vs. 8.5)) and also memory utilization. Important note is that the test is always using the same block, so after the fix it is being skipped every time. However conclusion can be made that the fix will improve performance in real life as well, since there are multiple submissions posted for the same block. In real life improvements may not be so evident though.