One of the interesting things about how we evaluate new technologies is how we so often begin by assuming a great new technology is going to replace its existing competitors. A little reflection shows this isn’t true, yet we make that assumption, time and time again. Today, we’re doing it with cloud computing, just as we did it with e-commerce 10 years ago.

10 years ago, many pundits, and an unfortunately large number of investors, thought that e-commerce was going to replace bricks-and-mortar retailing. After all, they argued, you won’t have to maintain a storefront, and distribution will be simpler, and you won’t need all those salespeople…

It didn’t turn out that way. Bricks-and-mortar retail is still alive and well. It is also true, however, that e-commerce has become quite a business. Some types of retail, such as bookstores, definitely have suffered. Particular kinds of e-commerce, for which the Internet is very well suited, have taken off. E-bay realized just how well the Internet could support a global flea market.

This is all pretty obvious stuff, but the amazing thing is we forget it almost every time an exciting new technology turns up. Despite historical precedent, we’ve heard people say, “Cloud computing is the future; the question is when to make the move.” They say that today, security may not be right, etc. But the assumption is that it will be someday, and that in-house data centers will eventually go out of business.

Not so fast. The assumption that the economies of scale of cloud computing will always trump the local data center is about as unproven as that amazon.com would wipe out every bookstore 10 years ago. In that case, it turned out that cost savings, and price competition, were not the key factors. What really boosted Amazon was its selection, and its diversification into other product lines. Remember pets.com? The sock puppet? Not such a success. The cost savings were not overwhelming by themselves.

So the right question to ask about cloud computing is not when to move everything to it, but which things it does well and which it doesn’t do so well. One thing it does well, for example, is to provide developers with nice test servers, for which they can pay for only to the extent they are used.

One thing cloud computing doesn’t do so cheaply is to host big servers 24/7/365 with lots of bandwidth used for traffic that would be local if the server stayed in the data center.

So maybe you keep the big custom apps in the data center, but you respond to requests from departments for development servers that they should put them in the cloud. They can also pay for them directly out of their budgets…

IT managers will find themselves doing triage like this very soon, if they’re not doing it already. Different projects will do best with different hosting arrangements. Get ready to make choices.

Of course, there are exceptions to every rule. The big apps might be best served in a community cloud, where a pool of expertise can be shared by similar organizations. But community clouds are a subject for a later post…