This one bugs me.
All the talk about Virtual Desktop Infrastructure (VDI) and how VDI is the next IN thing is beginning to look like hulla baloo to me. Every storage vendor in town is packaging their VDI messaging in the best gift wrapping paper possible, trying to win the hearts of potential customers. But I have a creeping feeling that the customers in Malaysia and even perhaps some in the region are going to be disappointed when all the fluff and huff of VDI meets reality.
I have to admit that I have no experience with VDI. I have no implementation experience, and I have no selling experience of VDI, but having gone through the years looking and observing at the centralized computing and thin client space, history could be repeating itself (again!). Many previous pre-VDI experiences have fallen flat on the face.
Remember the days of X-terminals, early versions of thin clients? Remember the names such as NCD (Network Computing Devices), Wyse Technologies (they were recently acquired by Dell), SCO Tarantella and the infamous Javastation? I don’t know about you, but that Javastation design was one ugly motherf****r.
So, it is my pleasure to remind you again and hopefully give you some nightmares too 😉
Back to VDI. Yes, the thin-client/zero-client/remote desktop/VDI concept is a great idea! I would have love VDI to be successful. It will be the implementation and the continuous user complaints that will be the bane of its problems. Ultimately, it’s the user’s experience that counts.
I was reading this document from Nimble Storage on their technology offering with VDI. And from their findings with the VDI experience, here’s a graphical representation of a typical day with VDI of several hundreds of VM:
The ONE thing that every storage vendor in town will talk about is the boot storm/login storm. Just look at the Read IOPS generated! As the day goes by, the OS updates and AV (anti-virus) scans also adds another degree of heavy Read and Writes IOPS.
The Nimble Storage whitepaper continues with an example of 2 workload profiles of a desktop user. Look at the table below:
And when you multiply by hundreds of desktop users, the I/O demands of an average VDI environment put a tremendous strain on the storage infrastructure. In the end, there is always (and I mean ALWAYS) the possibility that storage is to blame. Yup, storage is the culprit!
However, I tend to see it from another point of view. The storage technology and infrastructure can only do so much with what is available today. So is the network infrastructure.
After reading a lot of storage vendors’ presentations, listening to their talks, and reading VMware View 5 Desktop Virtualization Solutions book by Jason Langone and Andre Leibovici, a lot of considerations of the VDI solutioning have to go to the design and sizing of the storage for VDI.
And I have not started on the data protection and data availability aspects of storage infrastructure for VDI because that will be another big strain on I/O, despite what each storage vendor might say.
My point is, the whole desktop system is flawed from the beginning. The user interface of the windowing system is just too heavy. The desktop environment aims to deliver the user experience and to enhance the “user experience”, the programming toolkits, the widget sets, the whatcamacallit development environment are basically tying the need of the “user experience” to the fat client, not thin client. You need powerful hardware and heavy machinery to run those beautiful graphics, fluid movement, high resolution for the ultimate “user experience” and at the same time, you want this “user experience” on a thin client???
I even heard one guy from a trade show booth telling the crowd that their solution (not storage) could run MMORPG, high fidelity, full-HD graphics on their thin clients. Huh? You gotta be freaking kidding me!!!
Microsoft Windows (all versions included) and the X-Window System in the Unix/Linux environment are heavy stuff. The heavy desktop user interface coupled with the ever-demanding (or never-ending) user expectations of the acceptable user experience is just not the right thing for VDI. And it is this combination that has and going to kill the expectation of what VDI is supposed deliver.
Just look at all the thin-client solutions that were out there. Just look at how Linux tried to wrestle the desktop crown from Windows. Just look at X-terminals in the past being touted as replacement for high-end workstation for the Oil & Gas G&G engineers. Just look at ChromeOS. The success rate of these desktop has never been high, although Linux desktop push has been encouraging in the past couple of years.
Heck! We even complain when Windows runs poorly on our own desktops and laptops. Our “user experience” acceptance threshold is never satisfied. We continue to crave for the next best thing with the latest graphics card or the fastest processor yet.
So, in my conclusion, it is not VDI’s fault. It is US (aka you and me), and our damn “user experience”. Pretty soon, we will come to a point when we say VDI, VDI, VDI very fast.
Vee dee eye, vee dee eye, VDI, VDI, vdye! Email me if you get the meaning 🙂
Pingback: Say VDI very fast « Storage Gaga
I wholeheartedly agree with you “JavaStation” comment. I have used it during my earlier stint at SUN and despise it.
However it is possible to do VDI correctly and scale it globally. Schedule AV scans, patching and reboots over the weekend. Use citrix accelerators. Turn off all desktop eye-candy. For eg you really do not want the Aero interface on VDI. Host your VDI in a cloud. These are just a few techniques. As compared to a plain X-Session the RDP technology has a lot of tricks up it’s sleeve. For eg look at http://spice-space.org/
Hi Moinak
Thanks for your comments.
Yup, the Javastation implementation by Sun, even though it was a great idea, was a big challenge. Aesthetically it was just ugly. But even worse, was Java’s performance was bad during my stint at Sun in the late 90’s. These were early days. And imagine having this thin-client thingy over 100BaseT networks. 🙁
But what I am trying to point out in my blog are 2 things:
1) Our human expectations are always too difficult to be satisfied.
2) The windowing systems (be it Microsoft or X or others) are just too heavy to perform up to the satisfaction of human need for great user experience.
It’s getting there and I have seen SPICE in action. A lot of promise but RedHat and the rest of the open source world are still less marketable compared to the marketing engine of the likes of VMware. That also goes for the KVM battle with other commercial hypervisors. I read that KVM is just a dot in the whole hypervisor market share.
Thanks for your link, as well. All the best.
/Chin-Fah
When server virtualization started to gain momentum in 2007-2008, I thought it would go nowhere. From what I saw (been in the storage business since 1998), server vendors keep pushing the latest and greatest servers – faster and faster, with more RAM. How could they possibly say, not only do you not need to buy a new server this year, you could run 10-15 servers on one box!?
However, I was wrong, and by 2009 or so, virtualization was part of most projects I worked on, and is a must-have for IT shops today. In some ways, server admins and application owners had to learn to live with the new reality – your app is running on shared hardware (similar to how they had to learn to share spindles when applications were moved to shared storage some years earlier); the software used to manage, protect it, etc., are different; and other challenges that they might not have liked, but eventually came to accept. Over the years, the gaps have been filled – often by startup companies like Veeam, making software that not just mitigates the risk, but leverages virtualization’s advantages to make it even better.
With desktop virtualization, it is a different “customer” – the end user. From the developer who wants three workstations under their desk to do different tasks, to the bored slacker sitting in his cube looking for something to complain about, enterprise IT shops have a wide variety of people to satisfy. With server virtualization, I think what really helped was P2V’ing the easy ones – low IO boxes, test/dev, staging, etc. Once a comfort level was reached with that, then move up to higher priority systems, and now many companies are doing tier 1 apps. So I think that if one proceeds with caution, builds out an environment that can support a reasonable number of light users whose systems are going off maintenance and would be replaced anyway, a company can see some real benefit.
Over time, the percentage of VDI users will go up, but I don’t think it will hit the levels of server virtualization any time soon, so I agree with you there.
Hi Jason,
Thanks for your comment.
At the back of my mind, I am still a bit skeptical about the adoption of VDI in the users environment. You are absolutely correct that it will be spreading like wildfire like the server virtualization space. My argument when I wrote in my post is the notion that users are so fickle. The user experience and perception are the ones VDI has to contend with.
I am keen to see how the VDI space will develop, albeit with more user resistance.
All the best to you in 2013.
Regards
/Chin-Fah