0

I have a developer running automated tests which spike above 50% during setup, as seen in the below image.

He is arguing that the tests are unstable and difficult to troubleshoot, so downsizing is an inappropriate action. Ignoring the obvious response of "you need to fix your tests," is that even a fair comment? It looks like it's just spiking during setup, so if I cut this from an m5.xlarge to an m5.large is it going to see a stability change?

We have dozens of these being spun up at any given time, so the costs are becoming significant.

Load spike

SeanVDH
  • 101
  • 1
  • 1
    Setup might take longer, but as long as you don't mind that it does look like you're running at basically 1% of capacity after the initial spike. If you're running dozens at a time, just fire up one of the smaller ones and see. – ceejayoz Apr 30 '18 at 16:15
  • 1
    The beauty of cloud computing is, that you can simply try it out. Slower setup time could influence the stability at startup (especially with short timeouts) but it should not. – eckes Apr 30 '18 at 16:23
  • 2
    BTW the T1/2 micro instances with boosting are very good for that kind of bursty workload. – eckes Apr 30 '18 at 16:23
  • @eckes I'm definitely still getting familiar with AWS: is boosting something that I've got set up manually, or is it a default option? – SeanVDH Apr 30 '18 at 17:14
  • The micro instances allow a short time of higher cpu usage (or in case of T2 you can even configure it to alllow the paid option to use more for longer time) – eckes Apr 30 '18 at 17:19
  • Honestly this question may not have a black & white answer. I think the best course of action is to just try the various instance types and analyze the results. As others mentioned the "T" series instances are great for quick bursting rather than ongoing high utilization so they might be worthwhile to try. – emmdee May 01 '18 at 17:33

0 Answers0