I'm following this discussion with interest and I'm wondering what the goal of deltacloud is with respect to supporting every cloud model out there. The path that deltacloud is embarking on judging by the posts in this thread is to support the union of all features of all clouds. The current thread is about changing the machine characteristics of a server. I assume next is changing the characteristics while it's running. If I come along with a cloud that allows the user to change the color of the cooling fan LEDs can I get that added to the API as well? :-) What I'm trying to get at is that the number of dimensions along which IT has tweaked datacenter configurations is virtually infinite. Is the goal of deltacloud to express the superset of all the options out there? If not, what are the guiding principles?
The reasons for which I think it's important to keep the number of choices at bay are: - it quickly becomes a ton of work to implement any system that supports deltacloud: lots of UI to build, lots of rules to follow - the user is faced with a million choices, and we can easily end up in a situation where the majority of choices or combinations are invalid for any given cloud - doing anything in a portable manner becomes more difficult because the "conversion matrix" just explodes
In the end arguing over "I'd like 5GB instead of 4 and 3 cores instead of 4" is missing the forest for the trees. You win in the cloud because boxes are cheap, plentiful, convenient, on demand, and you can make them do what you need to quickly and in a scalable way. Not because you got knobs to tweak the last 20% of performance out of your favorite box. At least that's my experience with real customers spending >$10k/mo on EC2.
Sorry if this sounds like a rant, it's not intended to be. This is a tricky problem and I wanted to raise the question. Thanks much!
Thorsten - CTO RightScale