Can the rentable grid work?
Jonathan Schwartz is flogging the idea of the rentable grid - apparently, Sun is pretty close to being able to offer their $1/hour/CPU service within the US. I hadn't considered the export control issues with this before, and I'm glad I'm not on the management end of dealing with that one - sounds hard.
I do have a few skeptical thoughts though. A computing grid simply isn't - as Schwartz would like us to believe - like electricity. The power I use isn't any different than the power a factory uses (other than volume). Computing, on the other hand, varies. What about storage of whatever application is going to run on this grid? What about network latency? What about the language my application currently resides in? Unless I'm missing something, there are some fairly serious problems with just dropping my grid and using theirs.
Comments
Nope
[Byron] March 20, 2006 15:44:50.537
I'm a little skeptical. Universities have had pay-per-CPU services for a while and it never seems to work out all that well for the user (I personally have access to 2, but mostly use our group's tiny XServe cluster). I'm sure there are some users and probably some heavy users at that, but enough to be profitable? I think they got suckered by the focus groups: everybody always says that they'd LOVE to have access to something like that but in reality very few people actually use the service.