Open Stack

On Sun, Oct 14, 2012 at 5:24 PM, Sriram Subramanian
<sriram at sriramhere.com> wrote:
> Dear QA,
>
> As a prep for my brainstorming session on Performance and Scalability
> testing, I would like to know of any existing related work. Specifically,
>
> 1)  Do we have any instrumentation (preferably customizable) in place for
> any of the projects? Is it easy to enable for all the projects?
> 2)  How easy it is to add such instrumentation to Tempest?
> 3)  Do we have any metrics defined towards Scale/ Performance yet?
>
> I will be happy to discuss in person on Monday if you are already involved
> in such work/ are interested.

A small suggestion from previous scaling work I've been involved in:
test at an order of magnitude up from where you want to deliver: the
only sure thing is that folk will get load spikes, unexpectedly big
projects or busy projects etc, and having headroom becomes super
useful. You also find out bad algorithms that are taking considerable
time (but not enough to dominate profiles) - and those are sometimes
easy to fix, which makes the target size perform even better.

-Rob