[development] Question about Drupal performance

Jonathan Lambert j at firebright.com
Tue Jul 10 02:08:26 UTC 2007

All I know is that XEN doesn't crash, where as OpenVZ did suffer from a lot
of production stability problems.  I've run Virtuozzo, OpenVZ, XEN, VMWare,
LVS (which is good, and architectually sound).

Of those, XEN has required the least fooling and breaks the least of all of
them.  Performance is secondary to the brittleness many virtualization
technologies suffer from in regards to tuning various parameters - you want
something that "just works," and if you can afford it, it's hard
virtualization like XEN that delivers on that front.

I ran large scale hosting for eZ Publish (yea yea) on Virtuozzo and it near
about killed me.  Had SW-Soft telling us we were, "The most technically
demanding customers... [they had]," because we had all kinds of problems
with resources that would cause catastrophic failures of systems - I hear
its much better now.

I think the benchmarks I've done closely match what was posted (on XEN 3.03)
in so far as XEN really starts to shine when you have a higher load.  We run
a number of insanely high performance sites on XEN, including a couple of
clusters that use XEN for frontend web servers serving (in a couple of
cases) more than 1000 domains a cluster (some of which are very popular) on
Drupal.  And you know what the problem is with them?  MySQL.  Hehe, XEN
works flawlessly.

How did we get on this thread? ;-)

My 2c.

Jonathan Lambert

On 7/9/07, Anton < anton.list at gmail.com> wrote:
> On 10/07/07, Ashraf Amayreh < mistknight at gmail.com> wrote:
> >  I was actually disturbed when I saw this as my host is a VPS host that
> uses
> > OpenVZ. So I did a little search and I'm not so sure this statement is
> > accurate, not that I really care, but I'd rather be getting the best
> option
> > for my money so I thought I'd point this benchmark out.
> >
> >  http://community.livejournal.com/openvz/14024.html
> One thing I noticed about that benchmark (I only had a quick glance)
> is that while OpenVZ seemed to handily beat Xen on the low level
> synthetic benchmarks, Xen seemed to equalise or even slightly beat
> OpenVZ on the higher level benchmark (siege). You really need to
> benchmark your actual applications (eg MySQL, Apache, PHP etc) to get
> worthwhile data.
> Also Khalid probably used a more recent version of Xen than the one in
> the OpenVZ benchmark. They used Xen 3.0.2 which was a pretty early
> immature version of Xen 3.x. Xen 3.1 is supposed to be faster than
> 3.0.x was.
> --
> Cheers
> Anton

Jonathan Lambert
Principal & CEO
Email: j at firebright.com
Work: 415-376-7274

WorkHabit / FireBright, Inc.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.drupal.org/pipermail/development/attachments/20070709/b9743257/attachment-0001.htm 

More information about the development mailing list