Internet architecture: Not logical, captain!

Last week I recapped the results of some recent work I've done with my colleagues modeling Internet performance. In addition to assessing capacity and demand, we looked at what you could call "the logical Internet": the scalability of such protocol-layer features as addressing, routing, multihoming, and mobility.

The news there is pretty bad: Internet scalability is reaching its limits rapidly because of architectural issues inherent in the design of the 'Net. And IPv6 -- since the mid-1990s touted as the fix -- patently fails to fix the problem.

To see why, look closely at addressing. There are three types of names and addresses necessary for a complete architecture: application names, which are location independent and indicate what is to be accessed; network-node addresses, which are location dependent and route independent, and indicate where the accessed application is; and point-of-attachment addresses, which may or may not be location dependent but are route dependent and describe how to get there. 

To continue reading, register here and become an Insider. You'll get free access to premium content from CIO, Computerworld, CSO, InfoWorld, and Network World. See more Insider content or sign in.

Last week I recapped the results of some recent work I've done with my colleagues modeling Internet performance. In addition to assessing capacity and demand, we looked at what you could call "the logical Internet": the scalability of such protocol-layer features as addressing, routing, multihoming, and mobility.

The news there is pretty bad: Internet scalability is reaching its limits rapidly because of architectural issues inherent in the design of the 'Net. And IPv6 -- since the mid-1990s touted as the fix -- patently fails to fix the problem.

To see why, look closely at addressing. There are three types of names and addresses necessary for a complete architecture: application names, which are location independent and indicate what is to be accessed; network-node addresses, which are location dependent and route independent, and indicate where the accessed application is; and point-of-attachment addresses, which may or may not be location dependent but are route dependent and describe how to get there. 

A major problem with Internet architecture is that it names the same thing twice: Media-access-control (MAC) addresses and IP addresses both name the point of attachment, but there are no defined mechanisms for creating either network-node or application addresses.

Essentially, that means Internet architecture includes just the "how," not the "where" and "what." So what? This makes it incredibly cumbersome to implement such functions as multihoming (connecting to multiple networks simultaneously for load-balancing, greater performance or redundancy) or mobility (roaming across multiple networks). And IPv6 doesn't fix these weaknesses, it just throws a spotlight on them.

Take multihoming. In today's Internet, a URL first must resolve to an IP address, then to a well-known port. If a system has multiple interfaces (such as when it's multihomed), it has multiple aggregate-able IP addresses. The routers can't tell, however, that these different addresses go to the same place (because, again, there's no defined mechanism for doing so). So, the system has to be assigned a non-aggregate-able address, which increases everyone's router-table size. In practice, that means that either most users can't multihome, or routing tables must increase dramatically.

Or look at mobility. There are two ways to create mobility in today's Internet. One is for the user to stay entirely on a single provider's network, for example, a wireless provider's. That essentially makes the user into the carrier's captive. The other approach (which applies across, say, Wi-Fi and mesh networks) is to use IP mobility, in which the Internet creates a "home" router that knows when you move and creates a tunnel to the router where you are so it can forward your traffic to you. This clobbers performance: Imagine the impact on a voice call if it were being routed back and forth across the Internet.

In short, the Internet has some fundamental architectural flaws that are about to become glaringly evident, as billions more mobile and multihomed devices come online (including not just phones, but sensor networks and machine-to-machine links). As noted Internet Architect John Day says, "The Internet architecture has been fundamentally flawed from the beginning. It's a demo that was never finished."