[ovirt-users] oVirt 3.6 disk speeds

[Yaniv Kaul]

On Fri, Oct 23, 2015 at 6:55 AM, Julian De Marchi <julian at jdcomputers.com.au
> wrote:

> heya--
>
> Playing around with oVirt 3.6 on some new kit before I rack and stack in
> the DC.
>
>   - Dell Eql PS4210(Raid-6)
>   - Dell R630
>   - Dell N400 10GB switch
>
> I have my 10GB interfaces bonded and in a port-channel. I have setup a
> basic 3.6 oVirt self-hosted stack.
>
> I am seeing some pretty big differences in disks speeds from the
> self-hosted engine and a VM guest. Both are direct luns via iscsi from the
> SAN on the 10GB port-channel. Normal network traffic does not go over the
> 10GB.
>
> On the engine I'm seeing this:
>
> # dd if=/dev/zero of=~/testfile bs=1G count=1 oflag=dsync
>

You are not using direct IO, please add relevant flag.


> 1+0 records in
> 1+0 records out
> 1073741824 bytes (1.1 GB) copied, 1.58877 s, 676 MB/s
>
> On a test VM I'm seeing this:
>
> # dd if=/dev/zero of=~/testfile bs=1G count=1 oflag=dsync
> 1+0 records in
> 1+0 records out
> 1073741824 bytes (1.1 GB) copied, 22.0732 s, 48.6 MB/s
>
> I have played with the disk drivers for my guest VM, changing between
> drivers does not change the speed at all.
>

I assume both VMs have the same image format and drivers?


>
> What ideas do you folks have for me to narrow down this slow down? I don't
> really know where to start with this one. I also noticed that 3.6 does not
> yet have ovirt-guest-agent available.
>
> I know DD is not the best test for disk speeds, but it works for what I
> need right now, iozone will be next when I'm happy.
>

dd is by no means a tool for disk speeds. Specially not when you are
writing zero's. I suggest fio .
Y.


>
> Many thanks!
>
> --julian
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20151025/00b1ba39/attachment-0001.html>