Desktop grid is free - I’ll have that please..

I was in a meeting, we were talking about scavenging it seems to polarize the debate, just mention it and you sometimes get different reactions. I’m originally a Windows Server engineer, so standards, production, standard builds etc were deployed into my mind at a young age, so I’ll tell you what I think.

Using the desktops for grid is a great idea, typically you’re using pcs that are idle, so you are maximizing the IT capacity, the efficiency of the IT platform.  Here are my reservations:

• That the grid applications don’t interfere with layered applications - I can’t be rolling out Oracle clients, MDAC versions on demand which might break something
• That the desktop environment is seen as free - it’s not, each user is charged a support fee for each desktop, there is an energy cost, say the grid makes my pc go 7 hours at 100% cpu, there is an energy overhead, who pays for that? The end user? The application team? IT/Facilities?
• That a desktop can be rebuilt if it fails - can’t keep rebuilding a desktop their is a cost and disruption to the end user.
• What about application testing - when you test your Excel based grid application, what pcs do you test against? Base build? Is it ‘tested’ on a group of ‘production’ desktops on the production network?  It’s fine saying it works in development, but move that on to a support guys’ pc with lots of layered applications, in-house and bought in ones, with dependencies on Internet Explorer, Mdac, Data Access components, or even .Net framework versions, you can’t test for everything but a degree of due diligence testing is needed.
• What is the SLA on a grid application running on desktops?
• What happens when we virtualize the desktop?  We simply move the grid service to the virtualized session?
• The licenses aren’t cheap for grid, who pays?  IT and then sell back as price per cpu hour?
• One final point - what pcs are excluded?  The trader pcs?  They’re the fastest ones though, will IT exclude their own?  What’s the percentage lost through exclusions?