Open xiang90 opened 8 years ago
This does not fit well with go's benchmark system at all. We probably should write a long running normal func instead of using bench test suite.
We can change it to TestScheduling
. In this way we can still make use of Go's testing framework and profiling tools. We can define N iterations times via command line.
Is this project still alive? Or would you kindly suggest any other tools other than kubemark?
We want to benchmark the scheduling rate. The scheduling rate has strong relationship with the number of scheduled pods and nodes.
To make the benchmark work, we should fix the number of nodes and the number of already scheduled pods. And we try to schedule a pod under exact same for N times to get an average.
Also we want to learn about the rate vs number of scheduled pods too. This does not fit well with go's benchmark system at all. We probably should write a long running normal func instead of using bench test suite.