8 Comments

Summary:

Netflix said it would be the first customer of a new European internet exchange in New York. The news covers two trends — the growth of the European exchange model and the launch of an effort to bring that model to U.S. shores.

amsIX
photo: Om Malik

Netflix on Monday signed up as the first customer of a European-style internet exchange that is launching New York: the AMS-IX, short for the Amsterdam internet exchange. The news is not just that some pond-hopping internet exchange has a big name customer — it’s also the first internet exchange on the planet to adopt a new set of rules of standards from the Open-IX initiative.

There are two really important things going on here; the launch of a Netflix-led open internet exchange model called Open-IX, as well as the growth of the European internet exchange model here in the U.S. Both elements of this story could increase competition in the market dominated by companies like Equinix, Telx and CoreSite and perhaps lower prices for the buyers of bandwidth in the U.S. Those buyers could be everyone form Google and Netflix to transit providers like Level 3 or ISPs like Comcast or AT&T.

For the consumer, these costs are hidden, but making the market for bandwidth more efficient is something that benefits everyone in the long term.

The manufacturing floor of the internet

First off, it’s important to understand what internet exchange points (IXPs) are. Essentially they are data centers where providers of web services and connectivity exchange links. So Netflix might house an Open Connect content server in an IXP, where a few hundred feet of fiber connect its boxes to those of a local ISP or a transit provider such as Level 3. This is how your episodes of House of Cards make it across the web.

However, this system isn’t perfect. ISPs might charge extra to add more connections, and in certain markets there are veritable monopolies held by the big data center owners, such as Equinix, Telx and others. These data center owners can make the cost of running that fiber between servers significantly higher than in other parts of the world, where a different type of internet exchange business model holds sway:

news20130924-1

The chart above, from analyst firm Telegeography, shows that cross-connect pricing in Europe is significantly lower than in major U.S. markets. But in Europe, pricing differences might not just be from competition, it could also be a function of business structure. European internet exchanges are owned by their members, which changes the dynamic. In contrast, Equinix in the U.S. is publicly traded, and smaller companies providing exchange services are privately held, but still for-profit entities.

Regardless of the specific business model, having exchanges is good for internet providers and eventually consumers. A 2012 OECD report on IXPs details how they lower the cost of transit significantly. the report dubs IXPs the “manufacturing floor” of the internet where bandwidth is created and then shipped to other locales. As an analogy, this has a few flaws, but it’s hard to deny that exchanges, no matter how they are run, make internet connections smoother and more plentiful.

Analysys Mason chart on IXP growth.

Analysys Mason chart on IXP growth.

Open-IX aims to disrupt the market

Part of Netflix’s news on Monday was that this is the first Open-IX effort, which begs the question of what exactly Open-IX is. The release says that:

Open-IX is an initiative started by a group of Internet network operators, content owners and distributors, and other market parties, including Netflix, that encourages the development of the neutral and distributed Internet Exchange model in the US. The ultimate objective of Open-IX is to reduce interconnection complexity and associated costs in this country, which have been more complex and higher than in Europe, where the preferred Internet Exchange model is more common.

It’s pretty clear that Open-IX is trying to promote the more European business model and, in an interview, David Temkin, director of network architecture and strategy at Netflix, and the Chair of the Board of Directors of the Open-IX Association, confirmed those goals.

Building competition in the bandwidth market

Temkin pointed out the U.S. model of a private company providing a space for interconnections means that whichever company controls a market becomes a gatekeeper for what companies might be able to put their servers inside its data center. Places like San Francisco or Atlanta have one dominant provider that sets prices without a lot of competition.

You also have islands of networks where a few large providers have data centers, but they aren’t linked, meaning a customer might have to buy connections in a variety of data centers as opposed to buying in one or two and sharing them. New York is a good example of this phenomenon, said Temkin.

Temkin’s familiarity with the European model and his job as chair of the Open-IX Association stems from his day job as the guy in charge of Netflix’s networking, and Netflix’s membership in a European exchange. In North America, Netflix traffic is about a third of all internet traffic, which means Temkin has a vested interest in bringing more competition to the market.

However, he stressed that this isn’t just about Netflix, but about improving the overall bandwidth market for all players. “We could have gotten lower pricing from providers like Equinex if we just dropped this Open-IX idea,” Temkin pointed out. But then the growth of bandwidth capacity would slow. Temkin’s not alone in that idea. Frank Orlowski, the head of marketing for the DE-CIX Internet Exchange in Frankfurt, backs that idea up (as does that OECD study).

“There are two different development paths for interconnections; the one in North America and the one in Europe,” Orlowski said. “The European model turned out to create better growth.”

So now, the idea is that data center operators like Equinix and Digital Reality and others will join as members, alongside content companies and ISPs, in owning and operating exchange points. While the end result might lower prices for the U.S. exchange providers, it also offers an opportunity to open up the markets where one player might have had a dominant presence.

Plus, even if the revenue from cross-connection falls for these providers, it’s a revenue source that might dry up anyway, as the silicon for 100 gigabit ports gets cheaper. That means that, instead of buying ten 10-gigabit ports from a provider, customers could buy one fatter port that’s not likely to be 10x the cost, said Temkin.

The European invasion

Open:IX is still small-scale, but the growth of European exchanges has been a trend for the last couple of months. In September, the DE-IX said it would enter the New York market, while in October, the London Internet Exchange (LINX) said it would open an exchange in Northern Virginia. In Texas, Cyrus One proposed a statewide IXP in March — Cyrus One is also a supporting the Open-IX effort.

As for the AMS-IX that Netflix has joined, the exchange has reached agreements with Digital Realty, DuPont Fabros Technology, Sabey Data Centers and 325 Hudson to operate and build an Open-IX Internet Exchange in the New York/New Jersey area. AMS-IX plans to deploy distributed Internet Exchanges in the Chicago and Silicon Valley markets, which are currently planned to go live in respectively the first and second quarter of 2014.

So the European-style internet exchange is hitting U.S. shores. The result should be more competition, and perhaps cheaper transit costs for the giants of the internet. As consumers this may not ever result in lower bills, but more competition and lower pricing at all of the connection points of the internet can only be a good thing.

You’re subscribed! If you like, you can update your settings

  1. Equinix, not Equinex

    1. Stacey Higginbotham tomo Tuesday, December 3, 2013

      Yes, fixed almost immediately. You’re fast!

      1. It’s not fixed yet… fyi, btw.

        However, he stressed that this isn’t just about Netflix, but about improving the overall bandwidth market for all players. “We could have gotten lower pricing from providers like Equinex if we just dropped this Open-IX idea,” Temkin pointed out. But then the growth of bandwidth capacity would slow. Temkin’s not alone in that idea. Frank Orlowski, the head of marketing for the DE-CIX Internet Exchange in Frankfurt, backs that idea up (as does that OECD study).

  2. I don’t see how this is any different from the current model of privately owned data centers, aside from the recurring cost of cross-connects. You still need to coax the telecoms to have a presence in the building, and power and cooling still need to be provided. The cost savings in the cross-connect fees is a drop in the bucket compared to the price of backhaul transport and equipment.

    Unless I’m missing something, this looks like an answer to a question nobody asked.

  3. Also, it’s Digital Realty (instead of Digital Reality).

  4. Regulations in the last and mid mile (residential, commercial, special access, and wireless) has a big impact on concentration of exchanges. Europe has a different regulatory and business model than the US. Open-IX is a good start, but we need layer 1-2 open access policy applied uniformly (not just to telcos) in the last mile. Any provider granted a license to a public right of way (wired or wireless) should be held to an open access standard. Then we can forget about the farcical net neutrality concept and let the market develop of its own accord, as it began to do in the 1980s-90s. As we’ve discussed elsewhere, Google should have done that in their last mile experiment, but then they want to preserve their own market monopoly in advertising higher up in the stack. Opening the lower layers too quickly will potentially disrupt even them.

  5. As long as there are no monopolies on the new technology Open-IX is planning on implementing, this partnership and introduction into the U.S. market is very positive. The U.S. already lags significantly in terms of average bandwidth and this model will help reduce the stress that video is placing on data centers as a whole. There will always be a first to market, but in the U.S. competition will soon follow and ultimately benefit the end users.

  6. Decently run non-commercial exchanges are not unique to Europe. Look at SIX for one that has been running for 15 years and is quite successful. http://www.seattleix.net/

Comments have been disabled for this post