Virtualization Performance: Linux VMs on ESXi

We introduced our new vApus FOS (For Open Source) server workloads in our review of the Facebook "Open Compute" servers. In a nutshell, it a mix of four VMs with open source workloads: two PhpBB websites (Apache2, MySQL), one OLAP MySQL "Community server 5.1.37" database, and one VM with VMware's open source groupware Zimbra 7.1.0. Zimbra is quite a complex application as it contains the following components:

  • Jetty, the web application server
  • Postfix, an open source mail transfer agent
  • OpenLDAP software, user authentication
  • MySQL is the database
  • Lucene full-featured text and search engine
  • ClamAV, an anti-virus scanner
  • SpamAssassin, a mail filter
  • James/Sieve filtering (mail)

All VMs are based on a minimal CentOS 6 setup with VMware Tools installed. All our current virtualization testing is on top of the hypervisor which we know best: ESXi (5.0). We have changed two things in our vApusMark FOS setup: we upgradeded the guestOS from 5.6 to 6.0 and increased the number of vCPUs of the OLAP VM from 2 to 4. This small upgrade means that our latest results should not be compared to the results in our older articles.

We (Tijl Deneut and myself) tested with four tiles (one tile = four VMs). Each tile needs nine vCPUs, so the test requires 36 vCPUs.

vApusMark FOS

The benchmark above measures throughput. As for response times, let's take a look at the table below, which gives you the average response time per VM:

vApus FOS Average Response Times (ms), lower is better!
CPU PhpBB1 PHPBB2 MySQL OLAP Zimbra
AMD Opteron 6276 2.3 671 514 1410 758
AMD Opteron 6174 2.2 674 524 1210 861
Intel Xeon E5-2660 2.2 645 394 160 631
Intel Xeon E5-2690 2.9 362 288 40 483
Intel Xeon X5650 2.66 745 569 821 866

Considering that we may assume that the Xeon E5-2690 consumes considerably more than the E5-2660, it looks like the Xeon E5-2660 is the new virtualization champ. Let us check out the power consumption numbers under a realistic load.

Benchmarking Configuration ESXi Performance per Watt
POST A COMMENT

65 Comments

View All Comments

  • fredisdead - Saturday, April 07, 2012 - link

    From the 'article' .....

    'The Opteron might also have a role in the low end, price sensitive HPC market, where it still performs very well. It won't have much of chance in the high end clustered one as Intel has the faster and more power efficient PCIe interface'

    Well, if that's the case, why exactly would AMD be scoring so many design wins with Interlagos. Including this one ...

    http://www.pcmag.com/article2/0,2817,2394515,00.as...

    http://www.eweek.com/c/a/IT-Infrastructure/Cray-Ti...

    U think those guys at Cray were going for low performance ? In fact, seems like AMD has being rather cleaning up in the HPC market since the arrival of Interlagos. And the markets have picked up on it, AMD stock is thru the roof since the start of the year. Or just see how many Intel processors occupy the the top 10 supercomputers on the planet. Nuff said ...
    Reply
  • InsaneScientist - Wednesday, March 07, 2012 - link

    Johan, where in the specs where you have this line:
    Transistors (Billion) 2,26 2x 1,2 2x 904 1,17

    I sure hope that 2x 904 (Billion) is a typo... otherwise AMD has some serious explaining to do. ;)

    Should be 2x ,904 (I think? Would be 2x .904 for me, I assume you follow the same rules...)
    Reply
  • iliev - Wednesday, March 07, 2012 - link

    Page 5, Benchmark Configuration

    R2208GZ4GSSPP specs table... E5-2660 is 2.2Ghz, and not 2.9GHz
    Reply
  • dodge776 - Wednesday, March 07, 2012 - link

    Hi Johan,
    Always look forward to reading your server reviews at AT, but no SAPS benchmarks this time?
    Reply
  • ppennisi - Wednesday, March 07, 2012 - link

    For maximum VMware performance on Opteron Interlagos cpu under VMWARE it's better to disable C1E and enable, where available, HPC mode.

    I found myself on a fresh installation of ESXi 5.0 on Dell R715 that leaving C1E enable literally crippled vm performance.
    Reply
  • boudini - Thursday, March 08, 2012 - link

    I'm not sure I would recommend using iray as a reliable benchmark renderer in 3ds max. It is not a self configuring mental ray, but an unbiased renderer which behaves fairly differently to mental ray, and most other renderers such as vray, final render and brazil. It is comparible to maxwell and fryrender, but is very new compared to those two longer established unbiased render engines. It also attempts to use the gpu to add to its calculations as well - which could significantly skew results.

    Using mental ray or vray might well give you quite a different result, and besides I don't think iray is widely used in the industry.
    Reply
  • omega4711 - Friday, March 09, 2012 - link

    This. The results of iray are mostly dependent on the GPU. The lack of proper scaling certainly isn't due to Amdahl's law. Just use mentalray with small enough render buckets and you can easily satisfy 64+ threads.

    Also, due to the limitations of iray, it can (at this moment) only be used in about 1-3% of real world scenarios.

    Please, for all the people that care about these benchmarks, use mentalray and/or vray.

    Otherwise, it's a brilliant article.
    Reply
  • silverblue - Thursday, March 08, 2012 - link

    You've put that Interlagos has 4x2MB L2, but that would only be true for Valencia; Interlagos is 8x2MB. Reply
  • aranyagag - Thursday, March 08, 2012 - link

    you forgot the E5-2687W with a 150w tdp and higher speeds Reply
  • colonelclaw - Friday, March 09, 2012 - link

    Hi There,
    Thanks for an excellent article. With regards to the rendering benchmarks, would you consider using VRay as a rendering engine? It's fast becoming industry standard, is compatible with all the big hitters (Max, Maya, Softimage etc), is cross platform, and I believe, is incredibly well coded to scale with cores.
    It's also incredibly popular, not something you could say about iRay right now.
    Reply

Log in

Don't have an account? Sign up now