[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Xen-users] Xen Performance

  • From: "David Morris" <dmorris@xxxxxxxxxx>
  • Date: Sat, 6 Jun 2009 18:37:53 -0700
  • Cc: <xen-users@xxxxxxxxxxxxxxxxxxx>
  • Delivery-date: Sat, 06 Jun 2009 18:41:35 -0700
  • List-id: Xen user discussion <xen-users.lists.xensource.com>
  • Thread-index: AcnnCTu/HgkO1HT0Tq6Yh9Ff+PrcuwAB1edy
  • Thread-topic: [Xen-users] Xen Performance

My first observation would be that I don't trust any self measured performance values from a VM. There are tricky time usage allocation issues and I've seen and heard the 8% claims but didn't believe the folks knew how to measure the VM behavior w/o trusting the VM.

From: Peter Booth [mailto:peter_booth@xxxxxxx]
Sent: Sat 6/6/2009 5:44 PM
To: Luke S Crawford
Cc: xen-users@xxxxxxxxxxxxxxxxxxx; Amir Maqbool Ahmed; Javier Guerra
Subject: Re: [Xen-users] Xen Performance

Luke, All
On Jun 1, 2009, at 8:43 PM, Luke S Crawford wrote:

Peter Booth <peter_booth@xxxxxxx> writes:

Here's more context. The VMs weren't page scanning. They did show non-
trivial %steal (where non-trivial is > 1%)
These VMs are commercially hosted on five quad core hosts with approx
14 VMs per host and just under 1GB RAM per VM. Thats not a lot of
memory, but then the workload of one nginx and three mongrels per VM
is comfortably under 512MB of RSS.

I guess I don't know much about mongrel, but if someone was complaining to me
about performance of a modern web application in an image with only 1GB ram,
CPU would not be the first thing I'd look at.    

I look at everything. Yes 1GB is a limitation. The mongrel was configured taking that into account. 

so steal was >1%?   what was idle?  what was iowait?   if steal was only 10%
and iowait was 50%, I'd still add more ram before I added more CPU.

Theres no need to discuss hypotheticals. Lets look at real numbers at a busy time:

sar -W -f
00:00:01     pswpin/s pswpout/s
00:00:06         0.00      0.00
00:00:11         0.00      0.00
00:00:16         0.00      0.00
00:00:21         0.00      0.00

pswpin/s pswpout/s is equal to zero at all times, in other words, no swapping is occurring.
So disk isn't a factor here.

00:00:01        CPU     %user     %nice   %system   %iowait    %steal     %idle
00:00:06        all     84.42      0.00      6.92      3.08      0.96      4.62
00:00:11        all     92.46      0.00      6.15      0.00      1.19      0.20
00:00:16        all     90.24      0.00      6.37      0.40      2.00      1.00
00:00:21        all     88.42      0.00      8.98      0.00      1.80      0.80

We are clearly CPU starved. 

and his performance improved.  Disk is orders of magnitude slower than
just about anything else (besides maybe network)  so whenever you can
exchange disk access for ram access, you see dramatic performance

That is not the case. You will only see an improvement if disk access is a bottleneck

My point, however, is that Xen performance is not well understood in
general, and there are situations where virtualization doesn't perform

These sar readings on DomU do not tell the whole picture, nor do the studies that 
show Xen throughput is at worst only 8% worse than native Linux.

There are scenarios where the impact of virtualization on user response time 
can be a factor of 3 or 4.

This issue is poorly understood, has been seen and described in research literature, and 
until we get a handle on it and understand it, it will cause substantial problems.

With the increasing popularity of the cloud and virtualized environments, where there
is less transparency than a physical environment, we should expect that 
performance problems will increase.

Xen-users mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.