North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Re: Public Works Peering

  • From: Steve Gibbard
  • Date: Thu Oct 06 15:07:55 2005

On Thu, 6 Oct 2005, J. Oquendo wrote:

Now that I had time to marinate weird ideas even further, this is how my
previous idea `could` work for all parties. Of course those making
financial decisions would likely hate this idea since it would somehow
manage to "hurt" their business in their eyes...

States (or countries) would create a massive public NAP which would be
peered in each state. Guaranteed not to go down. Well 99.99999% (snicker)
guaranteed not to falter. This network would be funded by taxpayer dollars
and anyone wanting to peer would pay solely enough to maintain this NAP.
A few models to look at (based mostly on things I've heard rather than studying closely, so corrections are welcome):

Saudi Arabia -- Government run monopoly transit provider. Interconnects the licensed ISPs locally and provides international transit and content filtering.

India -- Government imposed manditory MLPA with paid settlements. Designed to convince VSNL (the monopoly international transit provider) to announce all their routes to all peers, but not having the desired effect.

Various other places -- Non-government MLPAs. Industry run exchanges, with an MLPA as a condition for participating. Often done through route servers. I think Hong Kong is the biggest example of this, with the route server announcing 13,000 routes. Really common in smaller exchanges in areas where there's huge (orders of magnitude) difference between transit and peering costs.

There are also a few exchanges without route servers, but where peering negotiation gets done on mailing lists readable by the other members, which looks very strange to my American eyes.

The non-government MLPAs seem to work reasonably well in some places. The two examples of Government regulation above don't appear to have led to significantly lower prices, usually the goal of peering. US networks tend not to like MLPAs because it reduces control, and do seem to be good at keeping prices down in the major metropolitan areas, so it's possible US peering coordinators are at least doing things in one of the possible right ways.