Netflix picks fight with internet exchange industry
Top exec criticizes high prices, marketing and partying
The internet exchange industry is ripping customers off, charging too much for features people don't need, and spending millions on staff salaries, unnecessary marketing and social events.
That's according to the vice president of network strategy and architecture for Netflix, David Temkin, who created a stir at a meeting of network operators in Chicago last week when he gave a presentation [PDF] titled "The real cost of Public IXPs."
In it, Temkin attacked a number of high-profile internet exchange points (IXPs) – including LINX, based in the UK, and Netnod, based in Sweden – for charging high prices while keeping their finances under wraps.
He noted that the organizations are making millions in profit, despite increased costs associated with non-technical activities such as marketing and social events. He also digs into several companies' financials while noting many of them simply refuse to provide details. Board members of LINX, for example, are paid $22,000 a year – a "pretty good hourly rate," he argues. Another big IXP, AMS-IX, actually made a $4.1m profit but listed it as a loss following "an accounting trick."
Temkin also attacked some of the governance models used by IXPs, specifically noting that some hold themselves out as "member organizations" but give members no effective say. "They can choose to adjust pricing whether or not their members give them permission to do so," Temkin notes.
Some of the profits are spent on useful things such as outreach and education, Temkin says, but much goes on "less altruistic" activities such as running meetings, "extensive marketing" including awards ceremonies, and "galas and social events." One slide featured the live music show at a dinner at a recent RIPE meeting in Copenhagen, sponsored by Netnod.
While noting that peering prices have gone down over time, Temkin flags that costs in other – more competitive – areas of the internet infrastructure market have fallen much faster. The clear implication is that money spent on high salaries, lobbying, parties and so on is the result of organizations expanding to fit their income rather than providing useful value.
"IXPs have not followed industry trends for consolidation," he notes, and have instead spread out horizontally. He argues that "it's time for cost innovation in the IXP infrastructure space," and questions whether "members" of IXPs should have some of their money returned to them rather than spent on dubious activities. "Should IXPs be co-ops?" he asks, suggesting that the current market of for-profit and non-profit organizations are looking after themselves rather than the internet overall.
Unsurprisingly, the presentation caused somewhat of a reaction.
Depending on who you talk to, Temkin's presentation was either a bold and honest criticism of an industry that remains largely opaque and unaccountable, or a poorly researched and unnecessary attack on specific companies that Netflix would probably like to get a better deal from.
Either way, the fact that he started the presentation – which you can watch online – by asking that people act respectfully when they ask questions and stay clear of ad hominem attacks, shows that the arguments have hit a sore point.
At issue is the fact that IXPs charge a huge disparity of fees. An online summary of fees shows the cost of a Mbps of peering varies from €00.104 to €00.475 ($00.118 to $00.537), depending on which company you go to. Temkin's targets sit at the top of the price range.
Those fees are not public, and in some cases companies are required to sign non-disclosure agreements before they are informed of them.
There is also little doubt that in the IXP market – like much of the internet infrastructure industry – easy profits have only partly found their way back down in terms of reduced fees. The predilection for seemingly endless conferences at different cities across the world has long been observed in the internet world. As has been the capacity for expensive social events and dinners at expensive restaurants.
Staff and board members across such organizations are also often surprisingly well compensated. And despite the internet's culture of openness, the inner functioning and finances of many of these organizations, which often put claim to a broader public interest remit, remain stubbornly opaque.
It is that industry-wide self-serving nature that Temkin is no doubt responding to, although many have been quick to point out that his criticism is undercut somewhat by inaccurate facts and figures.
Peering rates have been falling – in some cases significantly – and Temkin may not have accounted for that fully due to building his calculations around dollars and not considering the significant changes in exchange rates between Europe and the United States in recent years.
Temkin also looks at IXPs solely as providers of network exchange, where two that he specifically criticized – LINX and Netnod – are well known for providing broader services. For example, LINX runs a global public policy arm that lets much of the rest of the industry know about important policy changes that may impact it in political centers across the world. And in addition to supplying a network exchange, Netnod also manages one of the 13 critical root server operations on which the internet rests.
There is also the fact that while the IXP market is not fully open, there is still an abundance of companies that others can go to if they are not happy with their current operator. It is not a perfect free market but it is also not a monopoly or oligopoly. As with any market, bigger, more-established players have a tendency to use that fact to charge higher rates. Arguing against that is to argue against the realities of commerce.
What Temkin appears to be taking exception to is two-fold:
- That IXPs continue to see themselves as defenders of the internet acting in the public interest, when in reality they are acting increasingly like for-profit companies in their own self-interest.
- That the industry's continued lack of transparency is at odds with its claimed open nature and makes it harder to make good financial decisions on who to sign up with.
In a subsequent tweet, Temkin made his view known further: "'IXPs are really hard to value, so just trust me on it'--IXP executive. Everything that's wrong with the industry in one statement."
As for reaction within the industry, many have said Temkin's attack was unnecessary, while others have argued it will serve as a useful entry for long-needed discussions. A significant number feel it is both. ®
Sponsored: Beyond the Data Frontier