When development of the push service is substantially complete, we will need to obtain CPU and memory benchmarks (under normal operations and during usage spikes) to be incorporated into the service's Helm chart.
Note: We have used the https://github.com/thesocialdev/mediawiki-services-profiler tool for this with mobileapps and chromium-render, and it will likely be useful again here.
Results
Benchmark
Current service stabilized in 1500m core of CPU and 300MB of RAM.
- Processing ~188 req/s in avg and nearly 0% of failed requests
+------+-----------------+---------------+---------------+----------------------+-----------------------+------------+------------+ | Type | Name | Request Count | Failure Count | Median Response Time | Average Response Time | Requests/s | Failures/s | +======+=================+===============+===============+======================+=======================+============+============+ | POST | /v1/message/fcm | 383600 | 51 | 163000 | 168835 | 188.56 | 0.03 | +------+-----------------+---------------+---------------+----------------------+-----------------------+------------+------------+ | None | Aggregated | 383600 | 51 | 163000 | 168835 | 188.56 | 0.03 | +------+-----------------+---------------+---------------+----------------------+-----------------------+------------+------------+
CPU usage
Memory usage
Response times (ms)
Total Request per second
