互联网快完了！

In 2007, Nemertes Research released a dire report on Internet traffic. By 2010, it said, the "exponential" growth in demand for bandwidth will butt up against the "linear" investment in networking technology, and that whole Internet thing you've come to know and love will start experiencing "brownouts or snow days, during which performance will (seemingly inexplicably) degrade."

By mid-2009, this certainly seemed implausible. Millions of people now stream Netflix on-demand video to their computers and TV sets, YouTube has added high-quality options to its videos, and Hulu's launch showed that ad-supported Web video could be hugely popular. Despite the growth in video (which is usually pitched as the thing that will bring the Internet to its knees), "Internet snow days" were about as likely as real snow days in Havana.

Which is why it was surprising to see Nemertes President Johna Till Johnson double down on the doom-mongering in a May 2009 column for Computerworld called "The Internet sky really is falling." The article's point was largely that "we were right" because YouTube has "recently announced it's discontinuing video delivery to certain geographies due to—ahem—lack of access capacity." (We have no idea what's being referred to here, but it's certainly not related to anything happening in the US, Canada, Europe, and other places with good Internet infrastructure.) Also, ISP usage caps prove that there's some kind of bandwidth crunch and IP (both v4 and v6) are doomed.

But it's not just bandwidth brownouts and Internet Protocol that will destroy the Internet—turns out that network neutrality will, too. In an October 1 column, Johnson painted some new warnings on her apocalyptic sandwich board. Forget brownouts and snow days—what we're talking about now is the Internet "going away."

The article's headline pretty much sums it up: "Hello net neutrality, goodbye Internet."

The argument is that:

1. Internet growth rates are continuing to edge up by anywhere "between 50% and 100% year over year."
2. This puts strain on last-mile access lines (like your DSL or cable connection), and those lines are "excruciatingly expensive to upgrade."
3. Net neutrality means that you can't recoup these costs by charging, for instance, absurd rates for instant messages and cheaper rates for other traffic. And that means carriers will all have to "charge by the bit." And that will "have a domino effect on peering arrangements. Tier-one providers now peer for free with each other. Once they have no choice but to charge for bandwidth, free peering will go away. And one of two things will happen then—both unpleasant. Either user costs go up (to cover the costs of peering), or more likely, carriers won't bother to peer in the first place (because they can't charge users enough to recoup the costs of peering)."
4. When peering goes away, "so does the Internet—because you're no longer able to connect to anywhere from anywhere."

Poor Internet, we hardly knew you.

The reality

Fortunately, actual Internet traffic growth rates are between 50-60 percent year over year, not 100 percent, according to the authoritative MINTS project at the University of Minnesota. And in countries like Canada(where carriers revealed much of their data to regulators as part of a net neutrality hearing), growth rates have dropped from 53 percent (2005-2006) to 44 percent (2006-2007) to 32 percent (2007-2008).

The Internet's core has plenty of bandwidth, so traffic growth really poses the biggest problem for access lines. Fortunately, big gains in capacity in the last mile aren't "excruciatingly expensive." While Johnson's single example is the most expensive last-mile buildout in the US (Verizon's transition from copper lines to fiber optics), cable and DSL operators can upgrade their lines for bargain basement pricesby adopting DOCSIS 3.0 (cable) or by running fiber deeper into the network (as with AT&T's U-verse, which already offers 18Mbps connections over copper wire compared to 6Mbps on the rest of its network).

Even Verizon, which is dropping $18 billion on the job, is doing so in the very sort of environment that Johnson says will sink the 'Net—one where neutrality is assumed and differential protocol pricing is not utilized.

And what "charging by the bit" has to do with peering remains a mystery; peers choose to "peer" (exchange traffic freely) with one another because they exchange roughly similar amounts of data. Customer billing methods don't affect this decision.

Besides, data increasingly fails to pass through the big Tier 1 providers anyway. New research out today from Arbor Networks and the University of Michigan shows that "over the last five years, Internet traffic has migrated away from the traditional Internet core of 10 to 12 Tier-1 international transit providers. Today, the majority of Internet traffic by volume flows directly between large content providers, datacenter/CDNs and consumer networks. Consequently, most Tier-1 networks have evolved their business models away from IP wholesale transit to focus on broader cloud/enterprise services, content hosting and VPNs."

What's most odd about Johnson's argument about network neutrality is that she admits that this is default network behavior right now. And while she frets about the huge growth of Internet traffic, the reality is that the growth rates have been much faster in the past (doubling every year or faster)—and the Internet abides! As for ISPs not having the money to invest in enough infrastructure to keep up with demand, well… just take a look an ISP balance sheets. Tremendous profits are being made now, even as cable operators roll out DOCSIS 3.0 tech and boost download speeds to 50Mbps or 100Mbps.

In the end, the song remains the same: of course the Internet has issues, but some kind of network-killing "exaflood" hasn't materialized in two years and doesn't look about to wreak devastation on the Internet in the near future. What we have instead is declining traffic growth rates in mature markets, and big boosts to access line capacity (for Verizon and the cable operators, at least), plenty of bandwidth in the core—all on a network that has generally been neutral for decades.