Currently, the fake web-server pods scaled up (to a fixed limit) and down (to 1) at a fixed time interval. There can be various ways (mathematical functions) to auto-scale the pods to change the load behaviour for Prometheus to scrape from.
Using different scaling functions will help understand the benchmarking results of Prometheus in a better way as it changes the load characteristics that Prometheus is scraping.
Currently, the fake web-server pods scaled up (to a fixed limit) and down (to 1) at a fixed time interval. There can be various ways (mathematical functions) to auto-scale the pods to change the load behaviour for Prometheus to scrape from.
For example: a step-like function to continuously up-scale the pods until it reaches a threshold -- Example Commit: https://github.com/GoogleCloudPlatform/prometheus-test-infra/commit/5f0d371ec5896644d0db189cd172b75e542a0fa1.
Current function to auto-scale the web server:
The above commit (https://github.com/GoogleCloudPlatform/prometheus-test-infra/commit/5f0d371ec5896644d0db189cd172b75e542a0fa1) looks like:
Using different scaling functions will help understand the benchmarking results of Prometheus in a better way as it changes the load characteristics that Prometheus is scraping.