A Practical Navigator for the Internet Economy

GTE's Positioning of BBN, the Internet Peering Business Model pp. 1 - 8

Introduction: The first half of the November COOK Report focuses almost exclusively on peering - the technical and economic and sensitive political issues attached to one of the most critical and least talked about "fulcrums" on which rapidly changing Internet infrastructure is based. John Curran, CTO of GTE Internetworking explains how increased requirements are likely to drastically lower the number of fully peered backbones by the end of next year. Converstatons with those lower of the "food chain" show the development of the business model that has made possible the formation and growth of more than 30 national backbones during the post NSFnet era. We indicate some ways in which second tier backbones will have to evolve in order to survive. An interview with Randy Bush emphasizes technical issues involved in the current complex series of peering relationships

With WorldCom trying to swallow every player in sight, we interview John Curran whose new division GTE Internetworking has some surprises of its own in store. John explains BBN's decision to be acquired by GTE as being driven by the need for access to large sums of cash and to dark fiber in order to continue to compete as one of the handful major global Internet backbones. Since GTE has opened up its local service by creating a CLEC (GTE Communications) to compete with its LEC (GTE Telephone Operations),it is able to enter the long distance market and increase its already favorable cash business. With a 25% stake in Qwest fiber GTE also has access to plenty of dark fiber.

This situation leaves GTE Internetworking, as BBN is now known, with plenty of resources to be sold by GTE Communications and able to offer customers a seamless Internet solution that combines local and global connectivity. Furthermore GTE management has had the good sense to leave BBN (GTE Internetworking) intact as an independent business unit within the larger corporation. The internet operations that GTE had before the BBN acquisition have been folded into the Internetworking unit leaving BBN's management structure and operational culture intact - a very smart move on GTE's part considering the gulf that separates the world view of a top rank Internet provider from that of a phone company.

In the business model represented by GTE Internetworking the losers are the IXCs who, without the large cash flow foundation generated by local service customers, are competing against each other in a zero sum game to increase their long distance market share. We conclude that it will be difficult for these companies to generate the cash for the infrastructure build outs needed for the time a few years hence when IP networks eventually swallow voice traffic as well. In this context while WorldCom is structured so that under a single tight command it can easily do acquisitions deals, this single command will likely present problems in seamless integration of the players that it is acquiring.

John expects OC-12 to be backbone peering 'coin of the realm' by the summer of 98 and that fully peered nationals may be no more than about 10 to 12 by the end of 98. In view of these trends we asked Dave van Allen, owner of FASTNET, a major Northeastern regional, to comment. Van Allen largely agrees with most of what Curran says adding that he sees measured usage charging as regrettable but increasingly inevitable.

In the meantime as peering shrinks from 50 nationals down to ten or 12 there will certainly be an impact on the second tier backbones. We talked with a number of people to try to get a better sense of the consequences. In doing so we have gained a better understanding of a gradation in existing business models where one can multi-home like FASTNET to three or four major backbones and develop on one's own a strong regional backbone, to build outs where one mixes a percentage of peering obtained at exchange points with transit that one pays for from the big upstream providers. This latter business model enables would be national backbones to buy ATM based PVCs and build out national networks where the more exchanges they are at the more of the internet - including small regionals - they can peer with and the less they will have to pay for transit. The trick for these players is to be able to balance adequate income from down stream sales with the need for cash reserves as the cost of their transit expenses increases - something that regrettably seems increasingly likely in 1998.

Randy Bush on Technical Peering Issues pp. 9 - 13

Randy Bush, Director of Network Engineering at Verio, describes a range of issues of concern to the larger backbone providers. He notes that they have developed tools to help them check for consistency in the policies involved in managing large numbers of peering sessions. They find that public exchange operators need to work very carefully with all exchange participants to avoid such abuses as the ability of smaller peer to point default at a larger one without getting caught.

Among the big five new peering is now almost impossible to get. Moreover criteria for peering are generally no longer published. Many peering agreements can be terminated on 30 day notice. Many of the second five are large enough to begin to join the big five via private interconnections with some, but so far not all, of the big five. While some smaller regionals have suggested that the large nationals peer with them at local exchanges, Randy explains the technical and policy difficulties that doing this would entail.

He also explains the advantages of using loose source routing to debug operational problems at public exchanges. He points out that NAT boxes are getting, he believes, good enough to allow multi-homing to upstream providers without having Provider Independent IP space. He reports that conversations from those who measure traffic may indicate that SprintLink's traffic share has declined to the point where it is in third place after MCI and UUNET and ahead of GTE (BBN)

He offers a very interesting perspective on multicast which is needed to save backbones from the exorbitant bandwidth demands of applications like Real Audio. "Multicast hasn't taken off because it doesn't have the necessary architectural and protocol and business model support. It has been guru friendly and the gurus have played their games. This will change. But to make it change, you need a reliable protocol and BGP4++ will be it. Unicast keeps us plenty busy. We look at multicast and say oh my god, where's the income - in part because we don't know how to charge for it." There are other technical problems involved with multicast and public exchanges. Nevertheless, we are starting to see native mode multi provider multicast deployment, which will make the old approaches seem rather kludgy.

NANOG Conversations, pp. 14 - 19

Conversations in early September with Sean Doran and others point out that via NAT improvements we are gaining mechanisms "to extend the address lifetime expectancy not only of the IPv4 unicast address space in general, but of any given host in particular. That is, there are now mechanisms which can hide address changes from hosts that deal with address changes badly, while at the same time there is increasingly good software to assist with renumbering hosts."

A debate on Sprints' use of GSRs in its backbone expansion leads to criticism of UUNETs investment in Juniper and continued reliance on ATM. Finally at the end of September there was an interesting conversation between Sean Doran and Sean Donelan on the locality of traffic that backbones must deal with and consequently where aggregation should take place. On October 2 and 3rd Donelan also posted useful information about the validity of NOC contacts for 36 North American backbones.

A Look at the .us Domain, p.20

The .us domain is changing but it still is not commercially viable. Article to be concluded in December issue.

Editorial on NTIA Handling of the DNS NOI, pp. 21 -22, 24

After being established to deal with the "threat" that the IAHC effort on Domain Names was perceived to represent, the NTIA led task force is likely to endorse the IAHC process with some limited changes. This very strange change course in the direction of federal policy reflects a Brain Kahin and Mike Nelson policy draft that ignores much of the Congressional testimony from the end of September and misses the most fundamental issues of how to adopt domain reform with the least disruption possible to American stakeholders. Indeed, it is such a 180 degree turn in policy on the part of the administration that we wonder what could possibly motivate it. Noting that ITU secretary general Pekka Tarjanne has just stated in an interview that the ITU would indeed like to perform internet governance functions and might want to get involved in content, we wonder alound if the Administration believes that it can execute an end- run around Congress by asking the ITU to mandate key escrow encryptiion for member states who use the Internet.