March 26th, 2004

Internet Peering


internet-peering.PNGInternet Peering is one of the stuff that has been on the back of my mind. I have several long conversations with many people, such as Bill Woodcock and Bill Norton, over the last year trying to get a grasp of the issues. The problem is easy to frame but the solution is not so simple.

In a paper published by ITU-T SG31 (via ITU Newsblog), it says:

Interconnection in telecommunications will continue to be an important and difficult problem facing policymakers and regulators. But like many difficult problems, the solutions are not simple and cannot be neatly summarized.

Remember that Internet is is a networks of networks. These networks are connected either transit or peering. If you are a small network (e.g. home or office), you are likely to buy transit from a service provider (e.g. subscribe to a broadband, or leaseline etc). Your service provider would buy transit from a larger service provider and so on. But as you reach the top with all the Tier-1 ISPs, who do they buy transit from? They aren’t going to buy transit from each another…so they peer. They exchange their routes so that their customers can reach each another but (usually) no money exchanged.Theorically, anyone can peer with anyone. The Tier-2 ISPs can peer with each another and seriously, if you like, you can even peer your home network with your neighbor home network2. Now, would the Tier-1 peer with the Tier-2? Hahaha…of course not! Tier-2 are Tier-1 customers!

The problem occurs when a Tier-1 ISP, for example, Singtel in Singapore is barely a Tier-2 in US. And guess what? UUNet may be a Tier-1 in US but barely a Tier-2 in Singapore. So in US, UUNet don’t peer with Singtel (for obvious reasons) and likewise, in Singapore, Singtel don’t peer with UUNet! So when two machine in Singapore communicates, their data packets may be routed around the globe.

Network efficiency is just the tip of the iceberg. More problem occurs when group of ISPs work together and effectively cut others out of the market by refusing to peer. One of the main complains about the three local ISPs in what commonly known as the “Golden Triangle”3, that they effectively cut out the other competitors by refusing to peer.

To be fair, the three ISPs have no reasons to peer with anyone else since they are the only Tier-1 in Singapore. And that’s a valid and sensible business decision! I mean, you can’t expect them to peer with a Tier-2, regardless what Tier they are in their home country.

Yet on the other hand, many in the industry believes that this cosy arrangement resulted in the high transit cost in Singapore. But then again, the prices has a sharp drop (with some credit to open exchanges like SOX and Equinix) in the last 12 months so where are the market failure?

If these are enough, the content providers (like Yahoo! and MSN) also joined the peering game and and neutral exchanges trying to establishing themselves in this tiny island. As mention, the problem is simple, the solutions are difficult. But it is an interesting problem to think about…

Personal Opinion:

The peering/transit is a Power Game; where the strong vs the weak, where the strong gets stronger and the weak gets weakers; Negotiation power goes to the one who can consolidate the traffic.

1 ITU-T SG3 is the Study Group responsible for International Voice Interconnection. While there are some similarities between Telephony Interconnection and Internet Interconnection, a good friend of mine (who shall remain nameless because of his public profile) believe a new SG without legacy mindset would be better suited to tackle this then SG3. And yes, we both agree that ITU is in excellent position to handle this difficult problem. (Hope they are listening :-)

2 Not as simple of course. I mean, who runs an AS and BGP routing for their home network?

3 The triangle peering arrangement was setup in 94/95. It was an elegant solution to a technical problem we have back then.

Comments are closed.