Re: [Xen-users] Re: file-backed vs. block device performance

[Marco Mililotti <m.mililotti@xxxxxxxxx>]

Mark Greenbank wrote:
> Hi again,
>
> Since I've not heard anything I'm assuming that no metrics exist, right? 
> Would it be appropriate to do a small test? Is anyone interested?
> I was thinking of setting up and file-backed and partition-backed domain 
> and running lmbench to do some simple benchmarking -- does this sound 
> appropriate?
Hi Mark,
I've done some (basic) fs benchmarking on our setup:
- Sles10/Xen 3.0.2
- Xeon 5130 - 2Ghz
- 4 GB ram
- Disk: 1.2TB on Areca raid controller 1130 on raid6
- FS: xfs for xen VMs

Benchmark done using lmdd, command like:
Write) lmdd.linux of=... bs=1024k count=1000 fsync=1
and
Read)  lmdd.linux if=... bs=1024k count=1000 fsync=1
Commands are repeated at least 5 times to ensure no cache influence.

Here are some results:
VM has physical access to a logical volume
 * write -> 209 MB/s
 * read  -> 299 MB/s
I've done some tests using file images, too. Results where around 15 
MB/s (write) and 18 MB/s (read), If I remember correctly...
If interested I can give you more details.

With respect to native performance we have that the single VM was able
to reach the following % of native speed:
- write: 92%
- read : 88%
This was not intended to be a comprehensive test suite, but only to look
at some numbers wrt physical device utilization.

Regards,
--
-- Marco Mililotti
"Democracy is two wolves and a lamb voting on what to have for lunch.
Liberty is a well-armed lamb contesting the vote."
Benjamin Franklin

--
-- Marco Mililotti
"Democracy is two wolves and a lamb voting on what to have for lunch.
Liberty is a well-armed lamb contesting the vote."
Benjamin Franklin

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users