search

HPCE / hpce-2017-cw6

2 stars 17 forks source link

Time budget(s) for stress testing #27

Closed nicholzk closed 6 years ago

nicholzk commented 6 years ago

Would the performance of the stress testing use-case be evaluated over a range of increasing time budgets? If that is the case, could we have an idea of the ranges to be considered?

Alternatively, is there a single time budget we should optimise for?

m8pple commented 6 years ago

Good point - that wasn't explicitly specified, and the random ranges chosen are a bit shorter than I actually use for full testing.

My plan is to test for values in the range 2..180 seconds - I'd like to have done a bit longer, but rabbit starts to have odd behaviour for n > 2^40, and I think that keeps it in the range where things are still well behaved, even for fast implementations.

I'll update the spec to reflect this, thanks for pointing it out.