Virtualised XenApp Benchmarks – Does Anybody Actually Care?

I apologise in advance for YARP’ing (Yet Another Ranty Post) – truth be told, I’ve been biting my tongue for so long with regards to vSphere that I just don’t feel compelled to offer any VI3 based technical insights. So give me some leeway for the 5 or so days, things will return to normal after that. Well, as normal as they can be around here ;-).

So as my good mate Boche posted yesterday, we see yet another benchmark paper being sponsored by Citrix (albeit published by a reputable enough 3rd party, the Tolly Group) touting the superior performance of XenServer for XenApp workloads. I have 3 words for these kind of benchmarks. Who. Fucking. Cares.

And I’m not talking about the fact that yet again the benchmarks were performed on hardware that wasn’t EPT / RVI capable, or about how infintesimally small the performance piece really is in the overall virtualisation puzzle. I just can’t believe that Citrix keep pushing these numbers as if XenApp was the only thing that anybody ever ran on Windows, or was the primary use case for virtualisation. In case anyone from Citrix is reading this, NEITHER OF THOSE THINGS ARE EVEN REMOTELY TRUE. For one, XenApp boxes are in the vast minority of application servers in any infrastructure. Second, production XenApp servers are generally well utilised – they do not fall into the ‘low resource utilisation’ category which lead us down this path in the first place. And as such, it’s far from a crime against the environment to let them run a little longer on physical hardware. Besides, you may well conclude that VDI is a better option by the time you get around to virtualising those XenApp boxes, and just switch them off.

So where on earth is the logic in touting this single workload over and over again? I mean, if you really thought that people were going to see the results and then make a generalisation about XenServer performance, then why the fuck go to such lengths to ensure that such publications talk about the “optimised for XenApp” template? I bet most people don’t even know what that template does – certainly the types of people who would have the wool pulled over their eyes by these papers wouldn’t know. Oh and in case any of those people are reading this, all it really does is grant an abnormally large shadow page table allocation to VM’s that have this option enabled. Which is null and void on EPT / RVI enabled hardware. As I said before, funny how none of the benchmarks that show massive performance differences between ESX and XenServer have been conducted on such hardware.

In contrast, VMware or VMware sponsored papers cover a wide variety of workloads. VDI, SQL, Oracle, Java, Apache, Exchange, even XenApp… VMware doesn’t have any problems throwing down numbers on any workload. But the only benchmarks we hear Citrix crowing about are XenApp related ones. Even when they’re being published by VMware, Citrix crow about it. As if XenApp was the only workload on the planet that people were going to virtualise. As if XenApp was the final nail in the coffin for VDI, when in fact it’s the opposite way around.

So although I fully expect VMware to post something in response to this latest benchmark, they really don’t need to. Perhaps if we all just ignore this kind of study, they will go away. Not helping with that in this post though, am I!


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: