openstack team mailing list archive

[Oleg Gelbukh <ogelbukh@xxxxxxxxxxxx>]

Hello, Julien

Thanks for insightful numbers!

It really seems that having 1 node per zone for Swift cluster is
inefficient,
and it may be better to have more not so CPU-packed storage nodes with less
devices per node then few high-performance nodes with disk shelves.

--
Best regards,
Oleg

On Wed, Apr 18, 2012 at 12:56 PM, Julien Danjou
<julien.danjou@xxxxxxxxxxxx>wrote:

> On Mon, Apr 16 2012, Julien Danjou wrote:
>
> I'm adding a note about where the numbers come from.
>
> > Volumetry:
> > - 450 GB of storage used (each node has 19 TB)
> This has been measured with:
>  df -h /srv/node/c0d0p1
>
> > - 57 accounts
> > - 7929 containers in 7870 partitions
> > - 58158 objects partitions used so far
>
> This is what is reported in log:
>  Attempted to replicate 60 dbs in 0.29485 seconds (203.49118/s)
>  Attempted to replicate 7930 dbs in 27.41175 seconds (289.29204/s)
>  58158/58158 (100.00%) partitions replicated in 122.97s (472.95/sec, 0s
> remaining)
>
> Regards,
> --
> Julien Danjou
> // eNovance                      http://enovance.com
> // ✉ julien.danjou@xxxxxxxxxxxx  ☎ +33 1 49 70 99 81
>
> _______________________________________________
> Mailing list: https://launchpad.net/~openstack
> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~openstack
> More help   : https://help.launchpad.net/ListHelp
>