`networkperf` suite fails on `amd64` due to the inclusion of the `n4-standard-16` machine type

[kajiya3141]

This suite unfortunately fails altogether (i.e. no test results, just the daisy.log) due to the inclusion of n4-standard-16:

...
...
"[networkperf.create-vms]: CreateInstances: Creating instance "servervmt2dstandard1-networkperf-1tmz4".
[networkperf]: Error running workflow: step "create-vms" run error: googleapi: Error 400: [pd-balanced, n4-standard-16] features are not compatible for creating instance., badRequest"

so I pushed a branch to my forked repo with n4-standard-16 dropped and it does actually run the tests then.

However, there were still a handful of failures across TestNetworkPerformance over the different machine types, but curiously the FAILURE messages were: Expected: 85 Gbits/s, Actual: 94.7 Gbits/s (as an example). Which doesn't seem like a true failure if it's exceeding the expected target? In the code the if statement is if resultPerf < expected so I wonder if the output is being mangled perhaps?

[drewhli]

The test suite definitely needs some updates across the board. We haven't updated the target numbers either since the test was first introduced. The backend for CIT probably needs updating too to accommodate for different machine types, since looking at the error message, it seems like the machine type does not support the pd-balanced disk type.

[kajiya3141]

ACK, thanks for the insight @drewhli - I'll leave the issue open for now to track any future issues we may have when we come to re-include the networkperf suite 🫡