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 🙂