Benchmarks #2
Description
Are they actually useful?
What would be the most useful data to know: request throughput, average turnover time, data quantity served, concurrent connections?
What would be the most realistic scenario to test?
Example
Some preliminary results running on localhost.
Once all connections are established; all clients request index 5 files 0-4000 looped for a minute counting the number of responses they receive back without verifying the data's contents.
Specs:
i7-9750H @2.60GHz
| Concurrent Connections | Requests Served | Max heap | Min heap | Peak CPU (connecting) | Max CPU | Min CPU |
|---|---|---|---|---|---|---|
| 10 | 7786597 | 380352624 | 86797176 | 15.5% | 10.7% | 7.8% |
| 100 | 18822522 | 468273088 | 90785728 | 14.0% | 11.9% | 7.9% |
| 500 | 20427160 | 517946368 | 98670672 | 20.1% | 11.9% | 7.8% |
| 1000 | 20329146 | 381639840 | 100621472 | 21.6% | 12.0% | 7.9% |
| 2000 | 18378779 | 378485520 | 84875120 | 22.4% | 12.8% | 7.9% |
| 5000 | 13989759 | 580907712 | 111145664 | 34.5% | 12.6% | 7.8% |
| 10000 | 8777925 | 549515456 | 121230904 | 48.8% | 15.7% | 7.0% |
| 15000 | 6826589 | 543873920 | 153803648 | 60.0% | 11.3% | 5.2% |
All data was recorded in a static environment but mem and cpu data were taken with VisualVM so don't take it too seriously.