I sort of cheated since there isn’t a “Part 1” with this exact same title, but I have my reasons for changing the title and this is partly why I’m writing this second article about the same topic. The first article is entitled “Citrix Puts Storage ‘On Notice'” and can be found here. In case you haven’t had a chance to read it yet, I talked about the new PVS memory caching architecture and why it’s almost made the topic of storage throughput (aka IOPS) irrelevant. And there was some really good banter following the article’s publication – along with my “new best practices” article, it stirred up the most discussion via “comments” of any article I’ve ever written. And that’s obviously a good thing. But there are one or two comments in particular I wanted to address.
I <HEART> STORAGE VENDORS
Part 1 certainly had a provocative title. But let’s get one thing straight – I did not intend to alienate any of our friends in the storage space. And there is still a place in this world for many of the smaller storage players, too. I say this because a couple folks in the storage arena took offense to my “on notice” statement, and while I actually think that’s a good thing, I do want to set the record straight.
The intention of my first article was to educate folks about the amazing new PVS feature that debuted in 7.1 that reduces IOPS by as much 99% compared to pre-7.1 code. The breakthrough made by our PVS Engineering and LCM teams cannot be ignored. But the second reason I wrote the article was to make people think long and hard before they simply buy a 3rd party solution (or invest in expensive SSDs or some fancy flash-based array) to address the “IOPS problem” specifically. I honestly believe a few of the software-only players in that space will be out of business in a few years as companies like Citrix, VMware and Microsoft build better caching capabilities into the hypervisor itself. Even PVS should be irrelevant a few years from now – yep, I said it. Sounds weird, but it’s true – our customers desire simplicity and scalability with native tools, and PVS (even with this new amazing feature) should have a short shelf life if we continue innovating. This type of in-memory cache that can handle both reads AND WRITES should be implemented within the hypervisor. And if there is an array present, we can offload many of the “expensive” CPU tasks like thin provisioning or block-zeroing to the array (we already do this to some extent today – VAAI anyone?). Everybody wins in that scenario – simple, amazing performance with native tools – but we aren’t quite there yet.
MORE THAN JUST IOPS
As I was saying, we just aren’t quite there yet in my opinion. So PVS will be around for years to come. And many of our friends such as Atlantis and Nutanix can add value on top of Citrix, VMware and Microsoft SBC or VDI deployments. IOPS are just one problem we need to solve (albeit a difficult one and the most talked about…), but there are many other problems these folks can solve and PVS alone cannot address. What about persistent desktops? What about the IOPS associated with the pagefile and redirected AV DAT files and event logs? What if I need features such as snapshotting, de-dupe, quick backup & recovery or thin provisioning? Great questions – and many I addressed via the comments section of my last article – this is where our partners in the storage space thrive and make a lot of sense. But I stand firm – if you are thinking of buying one of these guys for pooled desktops or XA workloads to specifically address the IOPS problem, then I encourage you to test the latest PVS release and save your money. There is a time and a place for 3rd party storage solutions – IOPS is no longer one of them in my humble opinion.
PROFILES ARE NO DIFFERENT
And this is the same thing I’ve been preaching about profiles for the last decade. I’m sure I don’t have many friends left at AppSense and RES (and I pick on them the most because I love them and they are the most popular solutions I see in the enterprise space). But I would throw Citrix UPM in there as well. If you are buying and deploying these 3rd party profile solutions to address “long login times” or “profile corruption”, then I think you’re wasting money, especially if you haven’t tested a properly designed roaming profile solution with folder redirection of every shell folder, exclusions and Group Policy Preferences. I see so many customers still buying these 3rd party solutions (or even implementing our very own UPM) and they don’t even know why they bought them or what specific problem they are solving with it. Again, there is a time and a place for these 3rd party profile solutions, but I just wish people would think critically before simply defaulting to purchasing a 3rd party product because “we’ve had a lot of profile problems in the past”.
THE EASY WAY OUT
It’s the same reason I see many customers mistakenly choose XenDesktop or View versus XenApp (when all you need is XenApp to truly get the job done). And it’s why I see too many persistent desktop-based VDI solutions out there, too. I’m not saying there aren’t use cases that have to be addressed by XenDesktop or even persistent desktops with a fancy 3rd party profile solution – they do exist. But a lot of times our customers resort to VDI as opposed to SBC, persistent desktops vs. non-persistent desktops, etc. When I challenge even our largest customers in the world on a weekly basis as to why they chose XD over XA for use case ABC, or persistent desktops with Atlantis versus PVS pooled desktops for use case XYZ, or AppSense versus UPM (when you own it!), some customers don’t have answers and others don’t have the right answers backed by proper testing, education and research.
I am OK with 3rd party solutions as long as you know why you bought it and what business or technical problem you are solving. Hell – even Citrix is a 3rd party solution on top of RDS – so this is kind of like the pot calling the kettle black, but I encourage you to understand what value Citrix provides on top of RDS before you simply buy it because all the other Fortune 500 companies bought it, too. Do your homework and don’t take the easy way out! Spend the proper time up-front and you’ll have an extremely manageable and cost-effective solution long-term. (My buddy Dan also has some great commentary on persistent desktops if you’re interested.)
WRAP-UP
Anyway, I’m starting to preach and get on my soap box. But you get the point – you should not be buying a 3rd party product unless you have fully exhausted your options with the free, native tools (and you know why you’re buying it!). I love our storage partners that add value on top of Citrix solutions – I just don’t think you should be buying anything special to address the dreaded IOPS problem anymore. Times are changing and we should constantly be re-evaluating solutions and challenging conventional wisdom, as I’ve said in the past.
And by the way, on a totally unrelated note, I delivered a webinar last week to about 1000 people on PVS & MCS. If you attended, thanks for tuning in and I hope you enjoyed it. If you didn’t catch it, the recording is now live on Citrix TV and you can also check out the slides here. Thanks for reading and I welcome your feedback, as always.
-Nick
Nick Rintalan, Lead Architect, Americas
Citrix Consulting