A bit ago Brian Madden posted an article on LinkedIn proclaiming VDI had won the war against RDSH. I did respond there making the argument that is there is no winner or loser or even a war. The technologies are not mutually exclusive and will continue to coexist for many, many years. Shoot there are still fax machines out there that are in use today and I’m still waiting on the paperless office promised in the 80’s. BUT Brian does bring up a good point, adoption is on the rise of virtual desktops. Why is this though?
I actually had a very recent discussion around this very subject. I had brought up Brian’s article and told him of my response. He brought up his experience that companies were moving away from VDI and there wasn’t much of a future. I disagreed. I see a bright future for virtual desktops. Are there hurdles? Absolutely but I believe those hurdles are not necessarily technological anymore like they used to be but perception of older implementations failing to work as advertised. To understand why, let’s go back in time to when VDI started taking off.
When VDI first became a thing and companies started paying attention to it, the marketing machines at companies like Citrix went into overdrive when they saw a new market develop. In fact, if you think back, Citrix even tried eliminating the name XenApp altogether and call both XenDesktop. There were a couple of problems with this strategy though. One, it completely confused the marketplace and their own customers. XenApp was dead. XenApp wasn’t dead. There was a lot of confusion and, in my opinion, a lot of this confusion was sowed by Citrix themselves which brings up my second point. Secondly, during this time you wouldn’t hear anything about the RDSH side of things and it appeared that you could never get a straight answer or a consistent answer on what was going on with the XenApp technology. They were pushing VDI and trying to get customers to convert over to strictly virtual desktops. Remember the 2 for 1 licensing program?
The promise of VDI was you could replace your physical desktops with cheaper thin clients running sessions exclusively. It would be cheaper than managing physical desktops. You could run the desktop from anywhere. Many, many claims were pushed and early adopters found out that the promise of VDI, at the time, was not the panacea they were told it would be. Many early adopters got burned by this and the conversion would upset the users it was supposed to help. Now the term virtual desktops leaves a bad taste in the mouths of user in many of those companies. This makes moving forward very difficult with extending the reach of virtual desktops and some companies have given up completely.
How did this happen? If you think back to the early days of VDI, as a new and developing technology, there wasn’t a lot of knowledge on what would make a deployment successful. We knew we would need disk space and we knew we would need a certain amount of IOPs but how many was always a calculated guess. The biggest issue, at least in my opinion, is that we were dealing with SAN’s and in those SAN’s were spinning disk. In other words, very slow disk. There were ways of upping the IOPs but it was expensive to do so in a traditional SAN environment. SAN’s were not designed to handle the huge amount of disk activity that virtual desktops required. They were designed for server loads not a bunch of desktops. It became prohibitively expensive to push out a large amount of virtual desktops just from the SAN/disk perspective and users weren’t happy because their desktops were much slower than their old physical desktops.
Another issue that cropped up were the costs of the actual thin clients. They really were not much cheaper than buying a cheap desktop and managing those thin clients? Also not the easiest thing in the world. You also had to choose between Linux, Windows or a custom OS. If you didn’t choose a Windows based thin client, you ended up losing out on features of the protocol used to connect to the session. If you chose a Windows thin client, you then were back to managing Windows. Maybe not managing like a full blown Windows installation but you still had to do it. It was Windows after all in the end which also meant potentially putting anti-virus on those thin clients.
What this comes down to was that virtual desktops were either no cheaper than managing and running physical desktops or they were actually more expensive. This caused a lot of headaches for users, administrators and IT management who sold this internally to their company.
In the last few years though there have been MAJOR changes to many of these issues. The first change was brought about Hyper-Converged Infrastructure. Nutanix started in the End User Computing arena. They saw that traditional SAN’s were not going to be able to deliver a consistent user experience and developed a solution that could deliver a user experience that users would accept. HCI also made implementing virtual desktops solutions cheaper by combining storage and compute in the same solution. No more buying servers and then buying storage. You got both and it was so much easier to implement, manage and expand than the traditional way. HCI uses tiered storage including flash and spinning disks. The amount of IOPs that can be generated by these solutions are amazing.
As for thin clients, the prices haven’t necessarily dropped and you are still looking at Linux, Windows, custom OS’s or even zero clients but Citrix, VMware and others have drastically expanded the capabilities of clients that aren’t Windows clients. In other words, you have essentially feature parity between all the clients now. This gives you more options for moving away from Windows if you like and cutting down on management and other costs of those types of thin clients. If you decide to move to zero clients, it can potentially be even easier to manage which leads to savings.
My argument for the further expansion of virtual desktops is simply this: the underlying technology has finally caught up to many of the promises of VDI. HCI is a HUGE factor in making it affordable and being able to deliver an experience that a user will enjoy. Thin/Zero clients and vendor protocol clients have also come a long way making life easier for admins and technicians and even end users. Need a new device? Go to the closet, grab a thin/zero client, plug it, turn it on and you’ll be working in no time.
Now saying that, let’s be honest here. Some of those promises in the early days were completely unrealistic for the majority of companies. Being able to replace every physical desktop and go completely virtual? Not very likely. That one always bugged me and I feel bad for those that fell for it. The big problem for VDI now is the fact that people and companies did get burned by being early adopters before the underlying technology was available to make it actually feasible. Admins talk to other admins, companies talk to other companies and word of mouth can be both a blessing and a curse. Take into consideration users that have tried it in the past and had a bad experience and this can lead to difficulty to implementing or expanding virtual desktops.
One last thing I want to touch on and I believe this is going, and is actually proving to be, another big factor in virtual desktops adoption: high end GPU’s. Back in the day, putting a high end GPU in a server was not feasible or really supported by hypervisors. Then Nvidia came along and starting supporting projects like Xen to get there cards supported. VMware and Microsoft followed and now support GPU’s in vSphere and Hyper-v and the technology continues to evolve. Today a virtual machine can now migrate between hosts with GPU’s. A short time ago this was unheard of. A vm was tied to the GPU it was using. Why is a GPU important though? One, it allows many apps to function properly in a virtual desktop where as before it never would. ArcGIS, AutoCAD, etc can now be run in a virtual desktop and perform as if on a physical desktop. Second, even smaller applications can take advantage of a GPU making the user experience even better. Browsers, Microsoft Office and other applications will use a GPU if it’s available. The user experience is much richer when a GPU is involved. As with the other technologies the cost of implementing GPUs in servers is coming down.
Wow that’s a lot of words to digest but I really wanted to get this down and make the case for virtual desktops and why they are finally viable. Again, they are not going to completely replace physical desktops. It will never completely eliminate the need for RDSH. It does offer enormous benefits though when talking about accessing a users desktop from anywhere. The capacity to run complex applications that RDSH simply cannot or wouldn’t be a good idea to run in a multi-user environment. The ability to easily access information on the companies network without necessarily taking it off premises and now you even have the option of running desktops in the cloud.
Man, hopefully that all makes sense. Thanks for reading!