Open Stack

Hi Nick,

I'm planning to add rate-per-second meters deriving from the
cumulative disk and network meters:

  https://blueprints.launchpad.net/ceilometer/+spec/rates-derived-from-cumulative

However, the intent is to average these over the polling period,
which typically would be in the range 60s to 600s, and not to
incur the cost of sub-second hypervisor polling. 

Cheers,
Eoghan

----- Original Message -----
> Hi,
> 
> Well, that is the problem exactly - I need to get info every second to track
> traffic bursts, i.e. depth of precision should be on seconds level. So, in
> my case I`ll have to either poll it every second - or approximate it by
> computing an average…
> 
> Thanks a lot for your answers!
> 
> NM
> --
> Nick Maslov
> Sent with Airmail
> 
> 
> 
> On November 29, 2013 at 11:47:50 AM, Julien Danjou ( julien at danjou.info )
> wrote:
> 
> 
> On Fri, Nov 29 2013, Nick Maslov wrote:
> 
> Hi Nick,
> 
> > How do you think, is it a right way to meter such data with ceilometer? I
> > mean, such statistics should be transferred to collector each second, and
> > in
> > case of >1000 VM`s can easily bring down database. How do you think?
> 
> You don't _have to_ poll every seconds, you can poll every minute and
> compute an average. That really depends on what precision you need.
> 
> You can do the same things with the existing samples provided by
> Ceilometer.
> 
> --
> Julien Danjou
> /* Free Software hacker * independent consultant
> http://julien.danjou.info */
> 
> - signature.asc, 851 bytes
> 
> _______________________________________________
> Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
> Post to     : openstack at lists.openstack.org
> Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>