Stop Comparing the Internet to the Highway System

Although once apt, it is time to stop comparing the Internet to the interstate highway system. The reason is that when it comes to traffic congestion and performance for the average Joe consumer Internet user, today’s Internet is actually the inverse of the highway system. Here’s why. In the highway system most congestion occurs on major arteries, while local streets are usually free and clear—but on today’s Internet the opposite is true.

In the old days every Internet user was connected via a company, agency, school, or institution with a “good” dedicated connection. Some even had multiple connections for redundancy. Back then congestion occurred most often in the network core at peering/exchange points, or on trunks when a trunk went down (which happened often). The interstate highway model applied because each institutional user’s “driveway” was clear but the highways were often congested. 

In the early days trunks were unreliable compared to today’s fiber circuits. Furthermore fewer trunks led to less redundancy. Remember that the Internet was originally built by organizations that were NOT carriers. ISPs like BBN, UUNET, PSINet, and ANS leased circuits from carriers. Those circuits were enormously expensive and added a layer of overhead. 

Life is far different today. Most tier one ISPs ARE carriers, and the circuits come from another department in the same company. Recall how the carriers built out capacity during the Dot Com era, and the desire for spare capacity to deal with flash crowds and the like spurred carriers to richly trunk their networks. As a result, today’s average trunk utilization is miniscule compared to that for the early Internet.

Most significantly, the landscape changed dramatically as consumers flocked to the Internet in the millions. At first consumers casually poked around well-connected sites and used email. But now consumers are a major online economic force generating, not just consuming traffic, and constantly using the Internet in new ways. The genie is out of the bottle. Imagine the Internet with no consumers—just academics, corporations, and the government. No reviews of products on Amazon.  No posts on YouTube. No Facebook. No way!

Here’s why it’s now time to throw the highway analogy out the window. Access networks are highly oversubscribed. The consumer is on a shared access network with limited bandwidth and more incoming than outgoing bandwidth. This shared bandwidth limitation applies both to cable at the head end and to DSL behind the DSLAM. Cable subscribers share a limited slice of analog bandwidth with the rest of the neighborhood users. In the case of DSL, although the copper is not shared, the Internet uplink from the DSLAM is typically highly oversubscribed (in both directions but mostly down to the consumers).

Access ISPs peddle instantaneous bandwidth as high as the technology will allow between the consumer and the Internet core. But economics dictate that each consumer does not get a private line into the Internet core. It is a fact of life that access networks will be shared and oversubscribed. All of us can’t get maximum speed at once just like all of us can’t use the local subway at once. ISPs can deliver super speedy service under ideal conditions, but as consumers use the Internet more, the oversubscription problem gets worse and worse.

Imagine a telephone network that only allowed you to call companies and when you did call, you could hear the folks you called but they couldn’t hear you—and calling your neighbor was just plain impossible. Not a very useful service, eh?  Unfortunately that’s where we’re headed.

Better than the outdated interstate highway analogy is airline service. In the old days an airplane trip was a luxury. Today it is an overused, oversubscribed, commoditized service in which queues are the norm and frustration rampant. Internet access is moving in that direction—choked up at the airports but generally running smoothly once you are off the ground.

Stay tuned to our blog as we explore ways to steer us clear of a poorly performing Internet in the future.