Xen users - Aug 2006 - RE: Differences in performance between file and LVMbased images

Hi Alex,
> -----Original Message-----
> From: Alex Iribarren [mailto:Alex.Iribarren@cern.ch]
> Sent: 24 August 2006 16:27
> To: Roger Lucas
> Subject: Re: [Xen-users] Differences in performance between file and
> LVMbased images
> 
> Hi Roger,
> 
> I totally agree with all your comments, I also thought several of the
> results were very strange, which is why I redid the tests a couple of
> times. Actually, the original test was simpler (and I didn''t test
as
> many things). When I saw the results I was getting I changed a couple of
> things, like the synchronous write and the random reads.
> 
> The reason why I didn''t do a larger test was because I
didn''t have
> enough space for a larger partition, but I''m going to make a few
changes
> and redo the test.
> 
> Anyway, I guess this serves as a reference point for what we can expect
> when using relatively small files.
Absolutely.  Both sets of numbers (cached vs uncached) are equally valid in
their own context. It is just important that each set of numbers is put
forward with the correct description/explanation, especially in a Xen-like
environment when the caching in Dom0 may be very unpredictable if a large
number of DomUs are active and accessing the disk.
> 
> Roger Lucas wrote:
> > As one final question, what is the scheduling configuration for Dom0
and
> > DomU with these tests?  Have you tried different configurations
> > (period/slice) for the DomU tests to see if it makes any difference?
> 
> I don''t know anything about scheduling configurations, so I guess
I used
> whatever is the default.
I''m running Xen-3.0.2-2 and seem to be using the sched-sedf schduler (I
run
"xm sched-sedf" to see its parameters).  If I don''t set the
DomU to the same
period and slice as the Dom0 (-p20 -s15) then my network performance suffers
dramatically.  I haven''t checked if they have any affect on disk
performance
- hence my question.  Apparently there are better schedulers in later
versions of Xen, but I haven''t had a chance to upgrade.
> 
> Cheers,
> Alex
Best regards,

Roger


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