Web 2.0 has brought Sun's old slogan that The Network Is The Computer back into fashion, but few take it quite as literally as Cisco Systems. If Cisco gets its way, components like memory and processors will be linked by IP networks, not circuit boards.

Cisco outlined this vision while launching Data Center 3.0, a two-year push based on virtualization. The theory is that disk drives already have been moved outside the server, changing from locally installed component to virtual network resource. According to Cisco, the same will happen to everything else: If we can have storage area networks, why not RAM area networks? CPU area networks?

This is very much a long-term plan, not part of the initial Data Center 3.0 strategy. Long-term to Cisco means 10 years, which it says is within the life cycle of current products. So if you buy into Cisco's theory, you'd better buy networking gear that can handle and all this extra traffic.

And that doesn't just mean lots of bandwidth, at least according to Cisco (though, of course, it is happy to sell bigger and faster switches.) Network devices need to be smarter, too, able to look deep inside every packet and understand what's going on so that traffic can be sped up, slowed down, or blocked accordingly. These networked components also need to be combined into something usable. This is the role of Cisco's new VFrame DC appliance, the flagship Data Center 3.0 product. Cisco's intent with the VFrame box is to make the network programmable, essentially putting the newly networked resources back together again into virtual servers.

Behind all this is Cisco's search for something new to do. It dominates networking to such an extent that mere increases in market share can't bring growth at anything like the rate priced into its stock. It needs to grow the networking business itself, and its favorite (but not its only) way is to take a very expansive view of networking. People who wouldn't buy a Cisco server might buy a Cisco virtualized server resource.

Making the network do more is an obvious winner in the carrier market, where it fits with carriers' own desire to extracting value beyond simple packet delivery. AT&T doesn't want to be loaning out pipes any more than Cisco wants to be building them. It's also why Cisco initially came out so strongly against network neutrality. But does it make sense for enterprise customers?

As users decide, servers and switches will increasing overlap. Cisco also will increasingly compete with server vendors. The more server functions are outsourced across the network, the less there is for servers to do.