I'd like to pick up say the 90% and 95% quantile of latency and plot that against number of workers (I understand that's a test per point), so I can produce something approaching an SLA, i.e. I'd have the app meets this SLA on this hardware with this numbers of users at this sort of load and dies gracefull with more users
We've discussed showing SLA-type information and a plot like this seems like it might be a good way to do it.
From #58 by @johndrummond:
We've discussed showing SLA-type information and a plot like this seems like it might be a good way to do it.