Open Stack

On Tue, Jan 15, 2013 at 12:24 PM, Nicolas Barcet <nicolas at barcet.com> wrote:

> Hello,
>
> As a follow up to our last meeting and in preparation to the next, it
> would be nice to collect rationales on unit of meters in ceilometer.
> Here are a few random thoughts collected so far:
>
>  * Units should always be based on internation standard [1]
>

Where there is a relevant standard value, this makes sense.


>  * Units should be the same accross all meters making the same type of
> measurments
>

This would be convenient, but I'm not sure we need the rule. If we say that
measurements are taken in the smallest unit available (as Julien points
out, use bytes not kilobytes or megabytes), then we will have the most
accurate value we can. On the other hand, it doesn't make sense to measure
instance lifetime in nanoseconds, so there are bound to be exceptions
either way.


>  * Units should (or should not) be allowed to be modified
>

I have proposed that for grizzly we simplify by saying that all sources of
data for a given meter should report in the same units, and that the units
not change over time. This means the units reported in the incoming samples
are basically labels, but ceilometer won't look at their actual values or
use them to convert between units when responding to a query.


>  * Units should always be [integer|floats]
>

We can't say that values must always be integers, and I'm not sure what it
buys us to say they will always be floats.


>
> What do you think and what policy would you propose?
>
> Thanks a lot,
> --
> Nicolas Barcet <nicolas at barcet.com>
> a.k.a. nijaba, nick
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20130115/25f367af/attachment.html>