Just got back to my office from the BriForum (http://www.brianmadden.com/) show in Chicago. I was there working the Intel/Dell/Citrix booth as well as giving a classroom presentation. I wish I had more time to attend more of the classroom sessions but working the booth kept me rather busy. Here are my thoughts and observations from the show.
-No one size fits all.
-If it is not broke don’t fix it.
-The end user experience is important.
-IT shops will engage in desktop virtualization implementations for security, manageability and faster rollout of Apps, not so much just for cost savings.
-Interest in type 1 hypervisors is growing.
-VDI still being looked at first, before any other delivery model, but no widespread adoption yet.
-Desktop virtualization is far different than server virtualization.
During my presentation I had about 115 people in the room. I asked how many either had a POC in progress or were in production with vdi. 30-40% of the people raised their hand. Then I asked how many people had vdi in production with over 500 users. Approximately 6-8 people raised their hand. A lot of people are looking at vdi, but not many large implementations yet of vdi yet. At least, not of the people in my class that day. There was more discussion this year than last around other delivery models besides vdi. Organizations will likely have a mix of delivery models around desktop virtualization. In addition to vdi, os/app streaming, hosted/virtualized apps, local containers/vm’s are all part of the various delivery models that are available today. No one size fits all. No one delivery model is ubiquitous yet, and may never be.
Another comment I heard was, if you are happy with the way you currently manage your desktops, don’t change it just for the sake of change, or changing to some desktop virtualization model, thinking that it will save you money, it will not. Seems as though the word has gotten out that one needs to look at all the back end costs and software costs associated with implementing a desktop virtualization model, not just the end client costs. Some companies are moving to desktop virtualization models but for other reasons than just cost savings. Those reasons being more centralized management, security, and faster application rollouts.
It was interesting to see a small trend towards having the end point perform some of the application processing when possible, rather than having the back end servers perform all the cpu application cycles. Certainly all the type 1 hypervisor offerings perform all the application cycles locally but so now does Red Hat’s offering with the SPICE protocol. The SPICE protocol looks to the endpoint first and the server second to perform the applications. I think we will see more of this type of balanced approach in the future between the endpoints and servers. With intelligent clients we will be able to let the clients perform the application cpu cycles when conditions permit, and when they do not, the servers will do all the work and just send out the screens.
It was brought out many times that Desktop virtualization in not like server virtualization. There are many more things to consider like graphics, usb ports, user personalization, and sleep/hibernate modes that are not found on servers. So desktop virtualization, while not necessarily hard, can be very involved.
All in all, I thought it was a good thought provoking show. Brian Madden and his team did another excellent job. If you have never been to a BriForum show before, you owe it to yourself to go to next year’s show. See some videos of the show here. http://www.brianmadden.com/blogs/videos/archive/2011/07/22/brian-and-gabe-walk-the-floor-at-briforum-chicago-2011.aspx
Also some more info on desktop virtualization here. http://www.intel.com/assets/pdf/whitepaper/dv_understanding_desktop_virtualization.pdf