Who buys from whom?

Question

DrPeering –


Is there a resource that can help me understand what tier 1 providers serve which tier 2s and which tier 2s serve which ISP's?


For instance....who does Comcast buy from, who does Earthlink buy from  ?, how about all the LECs? and little cable-co's  Does a table exist somewhere that lays this out?


Chip

DrPeering –


So much is next to unknown. What are your predictions for 2013?


Arch Stanton

Answer

Chip -

There is no authoritative source highlighting the flow of money between ISPs, so people in the industry tend to make educated guesses based on a few sources:

  1. 1)The routing table. One can see clusters of networks being announced via an AS and this usually means that these networks are customers of that AS.
  2. 2)Traceroutes. As imperfect as these are, they do demonstrate an active path and many have attempted to infer customer-supplier relationships based on these paths.
  3. 3)Leveraging Social Networks. By social networks I don’t mean on-line postings, I mean interacting with people. I have found that bar conversations, conversations with suppliers and colleagues at operations conferences, etc. all provide anecdotal evidence and/or verification of the nature of the relationships between providers. It has helped me to paint the picture, fill in the gaps, or validate underlying assumptions in the models I have created.

Ultimately though, the relationship between two providers is a private one so it is difficult for those outside the relationship to determine if money changes hands. And since peering coordinators talk to each other for a living, these types of topics give peering coordinators something to talk about.

Some research work, for example “The Structure and Evolution of the AS-level Internet”  written by Amogh Dhamdhere (CAIDA) was presented at the Workshop  on Internet Topology and Economics conference in Georgia Tech. Mr. Dhamdhere’s deck highlighted some of the challenges using the easily available data, and some of the discussions afterwards centered around the problems with the data and the difficulties obtaining data from the operations community. I would suggest you review this deck and the others from this excellent workshop

Arch - here on New Year’s Eve are a couple of my predictions for 2013.

1) Content Providers will operate 1000-router IP networks.

Many content guys are deploying or talking about deploying top-of-rack ROUTERS instead of top-of-rack switches these days.

Many of the larger scale guys have maybe a 1000 rack deployment at their sites, so overnight they will have a network team that operates a multi-thousand node routed network! It will be at least within a data center, and probably interconnected with other data centers with their own 1000-router clusters!   Wild.  I heard of a couple of deployments where *each* of the top-of-rack routers might have towards 200Gbps uplink capacity!  Should make the router companies and their ecosystem investors very happy.

  1. 2)Remote Peering will grow in popularity.

This will be driven by the reduction in deployment costs (you don’t need colocation or equipment or remote hands for your remote routers) along with the willingness by eyeball networks in far away places to peer even if long distances are involved.

  1. 3)An IX “Keeping Local Traffic Local” will become a vague notion of the past.

Much like the 1990‘s notion that every IP device has its own globally unique IP address before the advent of Network Address Translation (NAT), the notion that exchange points keep local traffic local will be only one reason for an IXP to exist.

ISPs that land at an IX will do so as a part of their network footprint, to peer with those there at the IXP, and as a “Why Not?”, maybe peer with some local networks as well. This is overstating the point, but you get the idea - “local traffic local” will not be the dominant reason for an IXP existence.

With a fair amount of certainty I can say that the second edition of The Internet Peering Playbook will be selling as an iTextbook and as a newly minted ePub, and for the Amazon Kindle book. Fortunately the kinks have been worked out so that the images look sharp and are zoomable etc. Expect to see a chapter added in 2013 describing “The Business Case for Remote Peering” which supports both prediction #2 and #3.

  1. 4)Microsoft will do something interesting.

With Vijay Gill hiring away so much of the senior network talent in the industry, they have to be planning something of massive scale. It will be one of the more interesting network development initiatives to watch unfold in 2013.

  1. 5)Miscellaneous. I think you will see massive growth in SSD drives use in server farms, desktops and laptops and multimedia-rich HTML 5 based sites. HTML 5 will be wildly successful and IPv6 will be still gaining acceptance.

Advertising will continue to get more and more targeted and discussions about our privacy increasingly blurry. We will see linkages between loyalty cards and our TV watching behavior for example, so it will be easier for advertisers to tell what ads got us off of our seats and to the store to buy their advertised processed food stuffs. All of these initiatives help grease the path from what-we-see to what-we-want to what-we-buy. Soon it will form a complete loop so we in turn share what-we-bought and why-we-bought-it and how-great-it-is to who-we-know and the advertisers can just watch-what-we-do.

Happy-Holidays and....

Keep-on-Peering -

Summary