The Connected Web

Any cloud providers in the room presumably went several shades paler when a startup realized their worst fears at Friday's Under the Radar conference. Cloud server management startup Cloudkick demonstrated a significant new level of cloud interoperability reports ZDNet's Larry Dignan.

"Cloudkick ... rolled out a feature that allows customers to hop between providers — Rackspace, Amazon EC2, Microsoft and others — based on price ... If Cloudkick can navigate providers, instantly get the best rates and then provision accordingly the return on investment becomes obvious. Double bonus if it's all automated in a way where some algorithm finds the best prices and navigates cloud providers without user intervention."

Cloudkick isn't yet offering this capability commercially — the service is still in a free-of-charge beta and you can imagine this switching capability might be an obvious capability to include in a future paid version — but its software does already manage both Amazon EC2 and (Rackspace-owned) Slicehost instances. The ability to manage multiple cloud instances isn't unique; much larger and more established rival RightScale also allows customers to manage instances across several different providers from a single dashboard, and last week announced it will add support for Ubuntu Enterprise Cloud, which enterprises can use to run private or hybrid clouds. But Cloudkick is the first vendor to raise the notion of automatically switching based on pricing metrics.

This is exactly the sort of thing cloud vendors least want to happen as they edge their way down the path towards interoperability. If customers can switch instantly based on price, then clouds will be commoditized and the providers' margins will be sliced as thinly as salami. Of course one of the obstacles to making it work currently is the lack of comparability between clouds — there's no single agreed unit of computing, for example, so how exactly do you define who's cheapest? There may be other factors customers would want to take into account, such as service levels, track record, support contracts and so on.

I do believe, though, that what Cloudkick has hinted at will become concrete at some point in the future. There's too good a business opportunity — for example to provide 'follow the sun' computing, in which server instances virtually migrate around the globe to take advantage of spare overnight capacity in each timezone. Another factor to consider is that some applications are more mission-critical than others, and so for example it might save you money to run your expense management software on the cheapest cloud providers even despite the occasional outage, whereas you'll be prepared to pay more to host your core systems of record on a more stable platform.

These are all future scenarios of course, which will not play out until both the technology and the industry mature some more. But it does suggest that there's quite a bit of cost still to be squeezed out of the cloud computing environment before we can finalize ROI calculations. Or put another way, assume cloud computing's going to get a lot cheaper than it is already.