VDI: The Old Frontier
This is the final part of the VDI Saga I’ve been writing about on and off for the past several months. To catch you up, two years ago, I inherited a 1000 seat VDI infrastructure build in 2009 that wasn’t performing. The situation was a result of a trifecta of poor design, old (and slow) technology and VDI Sprawl. My solution was simple, pair down the active VDI desktops to a number the existing infrastructure could handle, leaving only student classroom and lab computers on the old system and then phase out VDI completely over time in favor of mobile devices for students.
In the last installment, I explained how we moved all our student VDI desktops onto a Nimble storage array which allowed us to limp along through the end of the school year. Over summer we made several changes. We setup a VMWare 5.1 instance using the appliance based VCenter server and migrated all district servers not associated with running View into the 5.1 stack. At the same time, we also moved these servers onto the Nimble Array. We doubled the amount of CPU and RAM in the View Database server. We increased C drive space on teacher desktops by 2GB (from 8GB to 10GB) and doubled the persistent D drive to 4GB, buying us some time until we could migrate them off the system and on to MacBooks. Then, with all student Virtual Desktops and district servers running on the Nimble, we fired the system back up.
Separating out the servers from the view desktops eliminated the problem of the View slow downs affecting the production servers. The Nimble took the load of both Servers and student VDI desktops well, although we were seeing CPU utilization on the array spike at 120%. However, we did not experience any negative performance that we could detect. Teacher desktops continued to run out of space and the old SAN continued to exhibit slow downs but overall the system was much more stable than it ever had been. With Windows XP running out of time, our Blade servers criminally low on memory and the Nimble array maxed to CPU capacity, getting Teachers off of the system was the next logical step to phasing out VDI. In December, that happened with the epic handout of 350 MacBook Pros to certificated staff throughout the district.
Looking forward, as more Chromebooks and iPads come on line, phasing out student Virtual Desktops will continue. Eventually all the backend VDI hardware, now going on five years old, will be shutdown and the legacy of VDI will be over. At least that was my plan. The new Director might have a different vision.
For me, I look at VDI as a cost neutral (at best) solution to a standardized office computing platform problem. For bank tellers where security and conformity are critical, it makes sense. For learning environments, particularly schools, not so much. VDI has a lot of moving parts on the back end. Unlike Bank IT departments, School district IT departments are generally understaffed. As a result, skills tend to be a mile wide and an inch deep. The many specialized skills (Database, SAN, Networking, VDI software stack) required to keep VDI humming along are usually in short supply. On top of that, fixed computers with one sized fits all desktops are not conducive to 21st Century Learning. Computers in classrooms need to be mobile, they need to follow the kids and they need to be adaptable, able to run what they need when they need it. VDI is a business solution to the student access problem. 1:1 mobile devices are an education solution to that same problem.
Locked down, one size fits all environments conspire to restrict users. Open platforms encourage users to experiment and learn. In education, we should be building open platforms for our classrooms and empowering users to make their learning experience with technology their own. But the IT part of me things a nice simple VDI in a box solution (with a nice fast Nimble Array) for the Office staff might actually be a good thing. What do you think?