NEW YORK ( TheStreet) -- For most of the last decade, computing has gone back to the 1960s.Companies including Google ( GOOG), Yahoo! ( YHOO) and Amazon.com ( AMZN) have been spending billions of dollars isolating their servers in specialized data centers. Much like the computer rooms of the "Mad Men" era, these cloud centers are entered only by experts. They're climate-controlled and scaled. As Google shows, they can be very beautiful. But the cloud is now becoming "productized," put into boxes you can house in a conventional server room. Piston Cloud even put a cloud stack onto a stick memory and now offers it for download. A start-up called Nebula drew a lot of publicity last month by announcing Nebula One, a private cloud system in a box. But within weeks it was followed by Hewlett-Packard's ( HPQ) "Project Moonshot", which offers pretty much the same thing.
This seems to go against the grain of what cloud computing is supposed to be, an infinitely-scaled virtual system. But in some ways it actually enhances that system by making elements of it redundant. You don't want your cloud services to go down if a rat chews through a wire in Amazon's Virginia data center, and with enough redundancy that won't happen. Again, this is something Google has been doing for several years, shipping smaller versions of itself to corporate clients and to telecom server rooms so that more requests are served locally. Netflix' ( NFLX) Open Connect system is geared toward providing the same redundancy for its media files. The private cloud is evolving into a mainframe replacement, drop-in hardware that saves money over the client-server system you have but doesn't necessarily require that you fire your IT staff. If you want to scale it, you may be able to keep it in the city by having your hardware live in mineral oil, Quartz writes. But the private cloud won't be a corporation's entire cloud. What most companies will wind up with are hybrid clouds, combining their own resources with those of public clouds, which remain much cheaper on a price-per-compute basis than what you can get with your own hardware. These systems will have extensive redundancy -- resources will live in several places at once -- so it becomes more like the Internet than any 20th century computer model.