In my last article, I mused on VDI—sorry, I mean EUC. What could have besmirched the fine name of VDI so much that it had to undergo a radical marketing overhaul? Yes, I know that EUC is much more than VDI, but most people still refer to EUC as VDI, especially outside of the ivory tower of Silicon Valley.
I also lamented the “year of VDI” and how every year since 2005 has been dubbed the “year of VDI.” This article delves deeper into that concept and explores why VDI has failed to deliver on its promise.
The thing about VDI/EUC—and this is something that the vast majority of people forget about—is that it is not primarily a virtualization play. Sure, Horizon View is a VMware product. Yes, XenDesktop is a Citrix product with a virtualization slant. Sure, the desktop functionality is delivered as virtual machines running on a virtual platform. However, it is not a virtualization lead. VDI is a desktop strategy; it is a management, policy, and process play; and at the end of the day, it is all about a single thing—and that is the end user experience. How long does it take for a user to log in to the environment? How responsive is their web browsing to them? How much time do their queries against their databases take (or more to the point, appear to take)?
I want you to remember that phrase: “end user experience.” Pick it up and put in a box on the shelf; you will need it later. It is for this reason and this alone that desktop virtualization projects fail. VDI was and often still is oversold—and inevitably, as a result, under-delivers. It starts with the salesperson entering the board room with their snake oil and bedazzling those within with figures of monumental savings. As Brian Madden said as far back as 2011, VDI is not and has never been a cost-saving solution. Yes, there may be cost savings, be they OpEx or CapEx, but they are not the raison d’être. It is about efficiency savings, reducing time at the deskside for support staff, reducing end point management by dumbing down the end point device, increasing the security of your outside assets by centralizing your data behind your bastions, and preventing your precious data from being stored on road warriors’ laptops that get left in roadside diners or the tops of their cars. This is why EUC makes more sense than VDI: it is the full package.
End user experience. Why do I say that is the most important metric? Simple: because at the end of the day, we, as IT personnel—whether vendor PSO consultants, system admins, or even architects sitting in our ivory towers—are just servants of the business and of the end user community.
End user experience starts with the migration of the user from the user’s legacy device. How did this go? Were their settings carried over correctly? Are their icons in the correct locations? Did their favourites carry over? Do all of their applications work as expected? And yes, do they still have that picture of Tiddles the cat as their backdrop? These things may seem banal to us as IT professionals, but to the user, they are very important. Their desktop experience is a part of their work experience. This is where tools like Liquidware Labs’ ProfileUnity, VMware’s User Environment Manager, or AppSense’s Environment Manager can help. Migrating profile management to these tools means migration of user settings is seamless. It smooths the migration process and increases the chances for a successful project. This all goes to aid a good end user experience, as in most cases, the migration process is your users’ first interaction with your EUC strategy. It is their first impression; if it goes well, your project is off to a good start. It is very difficult to claw back customer satisfaction when you have lost their picture of Tiddles; take this as read from somebody who has been there.
But migration is not the be-all and end-all of the end user experience. You have your ongoing usage, login and logoff times, application load times, and more. How do you monitor these? More to the point, how do you monitor and act before you have a Severity 1 issue in your environment? “What? A Severity 1 issue for a desktop? Get real!” I hear you shout. Now, let’s stop and think about this for a just a moment. Back in the day, when your desktops were physical, if you had an issues with a machine they affected a single user. Now, if you have any issues with a virtualization host that is overutilized due to a misconfigured or misbehaving application, you do not have a single user but maybe up to one hundred affected users—and they are all calling the help desk because their application is not responsive.
What you need is prescience—a bit of precognitive insurance. Or maybe just a canary in a cage. What exactly do I mean by that? Back in the days before gasometers, miners would carry with them into mine tunnels a canary in a cage. A canary is a lot more sensitive to the buildup of noxious gases like carbon monoxide than a human is, and it would die before the gas could kill the miners. Thus, it provided a warning to those working at the pit face.
What, exactly, has this history lesson got to do with the end user experience? What if you could have an early warning system for your desktop environment—your very own canary in a cage on each of your host servers, monitoring performance and response times and actively informing you of impending bottlenecks and potential car crashes? Well, now you can. This is what Login VSI’s new PI product does. It allows you to discover performance issues before your users do and enables you to proactively manage issues before they blow up in your face.
I can picture you all staring at the screen in disbelief. “But why, oh why, do I need another monitoring solution? I have metrics telling my what my CPU utilization is, or how many IOPS my desktops are driving per data store.” True, you do. However, these metrics, as good as they are, do not inform you if your users cannot log in or performance is not as they expected. They do not tell you how long it is taking for an average login. Login VSI PI gives real-world performance insights by simulating a real user and real user tasks. This virtual user logs in, launches common applications, and records how long it takes for the tasks to complete. The system then watches for any large discrepancies in expected results and generates alerts.
This is an awesome step in the right direction. To have these sorts of metrics is really useful. Metrics that are based on actual tasks, not nebulous performance statistics, are powerful stuff. For example, perhaps your hosts all appear normal, and your users are “successfully logging in.” However, they are never receiving their desktop. It is just at the dreaded black screen of loneliness. Traditional monitoring solutions will not register this. The first time you will know anything is wrong is when your users start ringing you. Login PI, on the other hand, can and will inform you. Imagine running this at 7:45 am every morning and finding this issue on one host. You can now preemptively remove the host or desktop pool from serving machines, thereby preventing a call’s even being raised.
Who knows? With such a tool in your armoury, your users may be so satisfied with their desktop experience they even invite you to their holiday party.