Cringely is at it again. As I have written previously, sometimes Robert X. Cringely seems to write well beyond his knowledge-base. Unfortunately, it seems to be getting more common. The recent bizarre googanoia column is like that. And in this most recent column, he makes several boneheaded mistakes and misstatements in the process of trying to make an otherwise interesting point about bit torrent and network economics.
Cringely badly misstates a number of facts in the process of making an argument about the notional value of bit torrent (the portion of the market for distribution of media that is already controlled by peer to peer bit torrent networks). In fact, on review, he may get every single important fact wrong. Which doesn’t mean that the idea that he presents is invalid or useless (although in this case it probably is). It simply means that we can’t get there from here. We cannot reach the speculative world that Cringely describes from the poor, fact-inhabited world we are stuck in.
And that is the problem: when smart, creative people speculate beyond their knowledgebase, they tend to arrive at fantastical, implausible but nonetheless fascinating places. We may be tempted to give such thinkers more credence than they merit, simply because their speculation is so fantastical and therefore so interesting? It’s the intellectual equivalent of a car wreck: we know it’s not good but we somehow can’t look away. Maybe this time we should.
In an effort to sort this out, let’s review the “facts” that Cringely cites and then ask the good people at AMS-IX to help us clarify them.
Cringely Claim: “The Amsterdam Internet Exchange (AMS-IX) is the second-largest Internet peering point after the mac daddy of all peering points in Seoul, South Korea.” .
Fact: If “peering point” means “Internet Exchange” then this is wrong. As I wrote previously, the KIX is more of a bunch of private interconnections in a place and has no public statistics. If we’re talking about density of private interconnections, the San Francisco Bay area and Northern Virginia both have to be seriously considered as extremely dense “peering points”, if the term is to have any meaning.
Cringely Claim: “AMS-IX peaks at 238 gigabits-per-second flowing between its 257 member organizations (ISPs and backbone providers, mainly) and as I am writing this the traffic is currently 225.8 gigabits-per-second, which shows the IX is running pretty close to capacity.”
Fact: This one is strange: equating a lifetime peak with total capacity, as though there is an implied constraint in the platform (rather than the demand). Since there’s no evidence that Cringely contacted anyone at AMS-IX for this story, I contacted them and asked a couple of questions. “AMS-IX is not even close to ‘capacity’ On the current platform we can easily handle more than double the traffic we are doing now.”, said Henk Steenman, AMS-IX CTO. Well, that settles it. Current peak is not equal to capacity. In a well-engineered network they are separated by expected growth and some measure of likely changes in traffic pattern over a planned time-period. The degree of separation is also governed by budget, of course. But to equate the two is silly.
Cringely Claim: ” My point in writing this is that the folks at AMS-IX have long believed that BitTorrent traffic represents about half of this burden but had no real way of proving it.”
Fact: I actually know a number of people at AMS-IX, as do most people in this industry. They’re not particularly secretive people. Henk Steenman (quoted above) is a prominent technical participant in IEEE ethernet working groups and a frequent contributor at technical conferences. Cara Mascini, Chief Marketing Officer for AMS-IX attends a huge number of international networking and peering events and is extremely well known. And frankly, at 250cm of height, Niels Bakker can be seen from here at my keyboard in New England. OK, he’s not quite that tall. But almost. Niels makes an impression. He’s also a technically prominent guy, having presented at the Chaos Computer Club conference last month, and scheduled to present at NANOG 39 next month. My point is that these are not hard people to find. And when you find them, you can ask them: ‘what percentage of your traffic is bit torrent and how long have you suspected this?’ The answer I got when I asked several people was “We don’t know, don’t care that much, and never said anything about this to Cringely”.
Cringely Claim: “Then in June of last year Swedish authorities raided piratebay.org, the world’s largest BitTorrent tracker, shutting it down for a few days. This shutdown instantly dropped AMS-IX traffic by a third or close to 80 gigabits-per-second. While thepiratebay.org is the largest tracker in Europe, it is far from the only one, so the idea that 50 percent of traffic is torrent related makes sense.”
Fact: Is it true that AMS-IX dropped instantly by 33%? Let’s hear from Henk Steenman again:
“In June of 2006 the peak traffic was around 150 Gb/s, so 1/3 could not have been more than 50 Gb/s. 80 is nonsense. The Pirate Bay was shut down May 31, 2006. Looking at the long-term graphs, there is a decline in peak traffic of the order 10Gb/s. I don’t know if that can all be attributed to the Pirate Bay issue. Last year’s summer showed the traffic to highly variable and most of this was attributed to the warm summer weather.”
It can be so difficult to make a good argument when the facts keep getting in the way. And by the way, the traffic stats for the London Internet Exchange (LINX) show the same flattening and variation during the summer.
Several AMS-IXers were annoyed by this piece from Cringely, and it might not be obvious why that would be. After all, they were getting publicity, and how bad can that be? There’s an element of someone talking about them and getting the facts wrong. That bothers most people, especially when the comments are attributed to being from them, and even more especially when technical people are involved. But I honestly think the biggest annoyance was the comment about them running near capacity. To Cringely (clearly a non-engineer) this was an off-hand comment to which he gave little attention. To some AMS-IX staffers, this was more of an insult. They were incensed that someone would imply that they were lax enough to run their network exchange at or near capacity. They pride themselves in their hard work and careful engineering, and they do not risk losing customer traffic to avoidable congestion.
Anyway, on to the actual subject at hand: bit torrent. Bit torrent traffic is huge and important. I’d like to know exactly how huge as much as the next guy. Bit Torrent may even be 50% of all network traffic. Maybe 80%. I don’t know, yet. And certainly from the data at AMS-IX, Cringely doesn’t know, either. Bit torrent is going to be an important force governing the development of edge and backbone networking over the course of the next several years and unquestionably many people are going to be tempted to limit or restrict it (please don’t shut it off until my current downloads finish, please). Renesys is actively engaged in work with a number of carriers who are trying to understand, and ultimately charge for, the different costs of carrying different kinds of traffic from different kinds of customers. Perhaps I will be at liberty to comment on that further in the future.
I’m actually a fan of speculative analysis and do some of it here. It’s an important part of predicting and understanding the big changes that are happening around us. But as we soar into the clouds of fantastical speculation and idle curiosity, let us do so from a modest springboard of fact, Mr. Cringely.