Daily image performance test
The AMI which is built daily should be tested for performance. We will only test the infrastructure overhead (e.g. time to run "echo hello world"). We should time the first job to run on a new machine as well as the time to run subsequent jobs. We should also time jobs running each of the available job images.
These metrics can be emitted in logs. A follow issues can push them into Cloud Watch or Prometheus so we can observe the performance on a dashboard.
Edited by Joe Burnett