Open Stack

> >> What are the performance differences with this approach vs
> >> options 1 and 2?  If performance is a major concern, then an
> >> event based approach seems like it could be more efficient.  Or
> >> is there a reason a poll approach is needed?
> >
> >Its not so much performance as timeliness - can we be sure that a
> >measure we need at 60s granularity will actually be taken every
> >minute on the dot, or with everything else going on within the
> >nova-compute service might we end up with lots of random jitter on
> >the measurement points.
> 
> I still like option 4, but keep it a separate daemon (like it is in
> ceilometer) to guarantee the timeliness.

OK, so that's an interesting thought.

Would the nova folks be accepting of an additional daemon, and the
implications for deployment, config etc?


> As far as "leakage" goes, we have heaps of nova "leakage" in
> ceilometer, I think we will have a lot less ceilometer in nova with
> this setup especially if there is some kind of common library that
> we can provide.

I guess I wasn't as concerned with leakage in the opposition direction,
on the basis that the measurer generally needs to know a bit about the
thing it's measuring, whereas the measuree doesn't necessarily need to
be aware of the logic that the measurer uses to decide when to measure
and how to interpret the results etc.

Cheers,
Eoghan