The IBM Labs team has been hard at work putting their new IBM® Flex System™ x222 Compute Node to the test with Citrix XenDesktop, and the results are both innovative and impressive.  But before we get to the numbers, it is important to understand that the x222 is far more than just a new blade server.  It is a high-density twin node server that is designed for virtualization, dense cloud deployments, and hosted clients. The x222 has two independent servers in one mechanical package; a double-density design that allows up to 28 servers to be housed in a single 10U Flex System Chassis.  And before you think you’ve seen this act before, let me assure you that the numbers below will show that both of the x222 servers are independent, capable machines, magically squeezed into a single blade.  And thanks to some very intelligent engineering by IBM, this has allowed for fast and smooth integration between the nodes, as well as intelligent fault tolerance and disaster recovery.

To get specific and technical, let’s look at the spec sheets for the IBM® Flex System™ x222 Compute Node:

  • Two independent servers in one mechanical package to maximize computing capacity
  • Powered by the Intel Xeon processor E5-2400 product family with eight-core processors with up to 2.3 GHz core speeds, up to 20 MB of L3 cache
  • There are 12 DIMM sockets in each twin node server, which support low profile (LP) RDIMMs and LRDIMMs, with a total capacity of up 384 GB using 32 GB LRDIMMs
  • Supports memory speeds of up to 1600 MHz
  • Support for 2.5-inch and 1.8-inch solid-state drives (SSDs)

More information see here: http://www.redbooks.ibm.com/abstracts/tips1036.html

And now let’s look at why I’m so excited with what these new blades can bring to the Desktop Virtualization market. For context, please note 3 things first:

  1. These numbers (and far more) are available in a newly published Reference Architecture written by our friends at IBM Labs : IBM Smart Cloud Desktop Infrastructure
  2. The supporting test for this RA were performed with: VMware® ESXi 5.1™, and Login VSI™ Medium Worker workload, utilizing Citrix PVS Server.
  3. The reference architecture actually covers both the new x222 and the lades, as the new and more traditional IBM Flex System x240 is also showcased. (Very clear guidance is provided as to when and why each blade type should be used.)

The Results:  The IBM testing shows that when configured with x2 E5-2470 2.9 GHz CPUs and 192GB of RAM, a single server within the x222 blade can host 152 concurrent users.

While that is certainly and solid and impressive number, and proves my early statement that each of the IBM x222 twin node servers are real and capable servers.  It does not really justify the title of my blog… at least not yet.  For that we need to play out the numbers a little further and use the power of the multiples.

1 x222 blade = two servers each supporting 152 VDI Medium task workers ….  Or 304 Medium task workers in the footprint of a single blade.

But let’s not stop there…

Each 10U Flex System Chassis holds 14 x222 blades… so, that means, drum roll please ….

4256 VDI users in a single blade chassis !!!

OK so maybe I have gone a little too far now, as the IBM reference architecture points out, you still need to plan for disaster and add room to run in recovery mode. And of course, your mileage will vary depending on your applications and user types.  However the power of the platform, and the innovation of the solution is clearly demonstrated here.  And the full RA does include very clear guidance to enable you to easily include your disaster recovery plans requirements into your deployment design.

And please don’t think I’ve given away all of the reasons to read the Reference Architecture.  Here are just a few more reasons to spend your morning coffee break with this the IBM Smart Cloud Desktop Infrastructure with Citrix XenDesktop RA in hand.

  • Should you upgrade to faster processors?  They have an answer… E5-2680 vs. E5-2690 with real numbers and a cost analysis
  • Is there such a thing as too much RAM?  They investigate that as well, and include an interesting tid bit about reduced bus speeds with certain high-density RAM configuration choices
  • What if you want to assign two virtual processors per users?  Read the RA… they did the testing and have your answers

Now, admittedly this Reference Architecture is based on XenDesktop 5.6, but I have it on good authority this same team is already working at putting together an update with XenDesktop 7 results, and I can’t wait to see what the Architects over at IBM labs come out with next.