openstack team mailing list archive

[Joshua McKenty <jmckenty@xxxxxxxxx>]

Greg - this is how we do it. See "Null-Tier Architecture" at http://www.pistoncloud.com/cloud-technology/cloud-architecture/ . Note that it's not just CPU - you end up needing more RAM and Network bandwidth as well.

Joshua


On Jan 8, 2013, at 8:22 AM, Greg Chavez <greg.chavez@xxxxxxxxx> wrote:

> 
> Razique,
> 
> Yes.  We have the OS on a RAID 1 and the rest of the disks are in a RAID 10. However, should we go with the Compute+Swift Node architecture, we'll be using this:
> 
> [root@kvm-cs-gen-09i ~]# df -h
> Filesystem            Size  Used Avail Use% Mounted on
> /dev/mapper/vg00-root
>                       130G  1.9G  122G   2% /
> tmpfs                 379G     0  379G   0% /dev/shm
> /dev/sda1             194M   33M  152M  18% /boot
> /dev/sdb1             559G   33M  559G   1% /swift1
> /dev/sdc1             559G   33M  559G   1% /swift2
> /dev/sdd1             559G   33M  559G   1% /swift3
> /dev/sde1             559G   33M  559G   1% /swift4
> /dev/sdf1             559G   33M  559G   1% /swift5
> /dev/sdg1             559G   33M  559G   1% /swift6
> /dev/sdh1             559G   33M  559G   1% /swift7
> /dev/sdi1             559G   33M  559G   1% /swift8
> 
> What do you think?
> 
> --Greg
> 
> 
> On Tue, Jan 8, 2013 at 11:04 AM, Razique Mahroua <razique.mahroua@xxxxxxxxx> wrote:
> Hey Greg,
> so if I understand well, you want to have the disk on the hypervisors be used as Swift nodes right?
> is there any underlying RAID?
> 
> Regards,
> Razique Mahroua - Nuage & Co
> razique.mahroua@xxxxxxxxx
> Tel : +33 9 72 37 94 15
> 
> <NUAGECO-LOGO-Fblan_petit.jpg>
> 
> Le 8 janv. 2013 à 16:28, Greg Chavez <greg.chavez@xxxxxxxxx> a écrit :
> 
>> 
>> We are in the process of replacing our Diablo KVM infrastructure with Folsom.  Up until now, our virts have been using the local Compute node's disk space for their images which, obviously, defeats much of the purpose of a virtualizing.  We are ready to implement an iSCSI SAN, but we're a bit bummed that we're going to essentially be wasting the copious disk space on all the systems we ordered to serve as Compute nodes.
>> 
>> SO... we were' thinking about doubling up our Compute nodes and Swift storage nodes, fully aware that this might require us to reserve more cores for the KVM host.
>> 
>> Has any one else tried this?  Is this clearly a bad, bad, bad idea?
>> 
>> Thanks.
>> -- 
>> \*..+.-
>> --Greg Chavez
>> +//..;};
>> _______________________________________________
>> Mailing list: https://launchpad.net/~openstack
>> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~openstack
>> More help   : https://help.launchpad.net/ListHelp
> 
> 
> 
> _______________________________________________
> Mailing list: https://launchpad.net/~openstack
> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~openstack
> More help   : https://help.launchpad.net/ListHelp