I was in a meeting the other day and we were discussing the benefits of virtual desktop technology. We are currently heavy users of virtualization at the server level and are now trying to understand the business and technical issues and benefits of Virtual Desktop technology, specifically Hosted Virtual Desktops. There were a number of very smart, tech and non-tech people in the room. But, during the course of our conversation, what struck me was the focus on the benefits of Hosted Virtual Desktops (HVD’s) and not on the hurdles that need to be cleared before we can really reap all the benefits that everyone is talking about.
OK, we all can probably articulate 3 or 4 HVD benefits right off the top of our heads, centered around control of the environment, flexibility at the client endpoint, and security. But, what about the hurdles? Desktop virtualization is still a pretty new technology. There was an interesting Wired article recently that you could look at as one of the historical threads of desktop virtualization (remember the old Network Computer?). We have come a long way from the days of the NC, but there are still some technical issues that need to be addressed before jumping in.
I am not going to get into the technical details, but I will highlight two areas (Servers & Storage) that need to be looked at within the data center. We can tackle other areas (Security, License Management, Provisioning) in a future update. But I thought I would jot these thoughts down to make sure that as everyone gets excited about HVD there are some realistic questions asked and calculations performed to manage expectations.
Server performance – if you are taking 100 desktops and putting them on 4 servers, you have to make sure the servers can handle this level of performance. It seems to me some folks miss the subtle difference between the current Citrix world – one instance shared amongst 100 users, and the HVD world where there are 100 users – each requiring their own session or instance. There is a cost to make sure you have the right server configuration and it is not necessarily obvious, especially if your IT management is used to the older Citrix model where one server could service 100’s of sessions, since they looked exactly the same. HVD’s ability to have individual profiles (among other things) means each session is not exactly the same. Companies coming from a Citrix mindset may have a little time adjusting to the different server demands.
Storage – Initially the math sounds great, if I have 100 desktops each with a 100 gig hard drive I can reduce that storage management nightmare by running those sessions on the enterprise SAN. But, there are significant IO issues that come into play. You are piling all those local IO calls onto the SAN, and there are channel limits that have to be calculated. There are some simple IO volume calculations that the SAN manager can perform from which you can actually back into the number of HVD’s you could run with your existing SAN environment. It is worth the calculation before getting into a pilot, since a pilot is not going to give you that production load stress test that you must do before you get your business partners all excited about HVD.
HVD is a great technology, and it is coming. In a Gartner report last year they anticipated the HVD market at less than $100 million in 2008 but at $1.9 billion by 2013. That growth is staggering. If we stick with the simple Server and Storage analysis above, the demands on the Server and Storage markets will be considerable. Having the right partners to ask the simple questions, and as with any new technology, having partners who “have done it before” is going to become more and more critical. I expect those partners to come from different pools in the distribution channel, since fewer and fewer vendors, suppliers or manufacturers can see across the disparate components of the data center that have these new architectural demands.