Message-Id: <>
Subject: NSFNET Backbone services after November, 1992
Date: Tue, 26 Nov 91 14:23:29 EST
From: Stephen Wolff <>
Status: R

On Friday, 22nd November, the National Science Board met and approved two actions proposed by NSF's Division of Networking and Communications Research and Infrastructure.

The first is a Project Development Plan for the continued provision of NSFNET Backbone services after the current agreement with Merit Inc. expires in November 1992.  The text of the Plan is attached below.  In numbered paragraph 3, reference is made to an "IBM-MCI spinoff"; this should read "IBM-MCI capitalized Merit spinoff".

The Plan has an agressive schedule, but even so does not show a new award until April 1993. Accordingly, as the second action, permission was requested and granted to extend the current Cooperative Agreement with Merit at approximately the current spending rate ($10m per year) for a period not to exceed 18 months.

Those - quite literally too numerous to mention - whom we consulted, beginning with the first Harvard Workshop over a year ago, who cared enough to listen, comment, challenge, and debate, we gratefully thank. All of you helped inform our decision and shape the Plan.

Project Development Plan

Continuation and Enhancement of NSFNET Backbone Services


This Plan fosters growth and competition in the business of networking while maintaining the stability and reliability of a service that has become a valuable tool of the US research and education enterprise. It also provides for enhancement of the Backbone by allowing vendors to offer services based on emerging digital offerings of the telecommunications industry. The duration of the Project is three years and involves two concurrent  solicitations, under one of which multiple awards are contemplated. Based on costs of the current Backbone, the three-year cost of the Project is estimated to be $18 million.


The current NSFNET Backbone interconnects sixteen nodes and is operated by Merit, Inc. under a competitively awarded five year cooperative agreement with the NSF. Connected to each of the sixteen Backbone nodes are one or more "resource centers" such as a supercomputer center or  a  national  laboratory, or regional networks (e.g., SURANET, CERFNET) which aggregate network traffic from scholars and scholarly resources  at academic, industrial, and government campuses. Regional  networks are autonomous entities, supported by their campus clients and, in  many  cases, by a subsidy from the NSF. Although they are, collectively, in a state of rapid change and growth in clientele and traffic,  their existence and  support is not at present an issue. The NSFNET Backbone is the only government-sponsored source of non-mission- restricted trans-national  connectivity for the scholarly community; this request to the National  Science  Board  concerns continuation of this connectivity after the cooperative agreement with Merit ends.



1. Emergence of competition and maintenance of stability

When the competition for management and operation of the current NSFNET Backbone was conducted in 1987, the ARPANET operator (BBN) was the only organization with experience in operating a nationwide network using the (now standard) Transmission Control Protocol/Internet Protocol (TCP/IP). Indeed, there was widespread skepticism that the winner of the competition, Merit, and its joint study partners IBM and MCI, would be successful in the NSFNET Backbone enterprise since none of the three had any TCP/IP experience.

Subsequent events proved these doubts unwarranted, and Merit's success, by triggering rapid and sustained growth in number of users and in usage, catalysed the emergence of  new private enterprises offering national-scale TCP/IP networking. Within the networking community there is broad consensus that, in "recompeting  the Backbone", the NSF must build on and sustain this new diversity of competitive TCP/IP offerors.

Hundreds of thousands of researchers, students, and other scholars  -  including many engaged in "mission-critical activities" sponsored by agencies such as NASA and the Department of  Energy - depend on the NSFNET Backbone and system of regional networks for uninterrupted, reliable  service every day. This community's natural concern for stability in the provision of national networking services presents, to a degree, a countervailing force to the pressure for competition and multiple providers discussed above.

The challenge to the Foundation is to construct a continuation of Backbone services so that the two worthy goals, stability and competition, are both fostered to the greatest extent possible.


2. Fair competition

In September, 1990 (the third year of the cooperative agreement between Merit and the NSF), Merit subcontracted the management and operation of the NSFNET Backbone to a  new  not-for-profit concern capitalized by IBM and MCI. There is substantial agreement in the networking community that, while providing for continued Backbone services, the NSF should assure both that the incumbent is not favored and that there is  an  equitable opportunity for other firms to participate in the long-haul TCP/IP networking business.


3. Timing

The complexity of these issues has been compounded by their timing: a credibly competitive TCP/IP networking arena has only arisen within the past two years, and became an urgent issue with the September 1990 IBM-MCI spinoff.

In the past year, NSF has sponsored and participated  in  several workshops  and  meetings, and has consulted affected communities, networking experts, and representatives of other government agency networks in a variety of other forums. Only now is this process leading to an emerging community consensus on the future of the Backbone.

In August, 1991, the Federation of American Research Networks (FARNET), a trade association of regional networks that use the NSFNET Backbone for trans-national connectivity, organized a workshop under NSF sponsorship to consider the future provision of  Backbone services.  Their report(1) affirms the need for continued strong NSF support for top-level Backbone services and recommends a recompetition during Fiscal Year 1992 with multiple awardees.

In the early Fall of 1991, the Networking and Telecommunications Task Force (NTTF) of EDUCOM which represents academic campus networks and computer centers met and issued a report(2) on the same  subject. They say "Uncertainty prevails because of the expiration...  It is imperative that... NSF take immediate steps to clarify their  intentions with respect to the stability of backbone services", and later strongly recommend "A   new, competitively awarded cooperative agreement" for continued Backbone services.

The Division Advisory Committee (DAC) for Networking and Communications Research and Infrastructure met on November 4 and 5, 1991, and considered all currently known options for the post-1992 Backbone. They overwhelmingly preferred a recompetition with multiple awardees.


4. NREN involvement

Further complexity has been introduced by the five-year High Performance Computing and Communications (HPCC) initiative in the President's fiscal year 1992  budget, which gives NSF the responsibility for implementing the National Research and Education Network (NREN) and coordinating the participation of other federal agency networks. Since the NSFNET Backbone will be a central feature of the NREN, the management of acquiring its services post 1992 is complicated by the need to treat the NSFNET as part of a total national information  infrastructure for the support of research and education, and by the necessity of multiple agency coordination.

In order to help meet its NREN responsibilities, the Division of Networking and Communications Research and Infrastructure engaged an independent engineering group (the NSF NREN Engineering Group, or NEG) to advise on technical matters of the implementation. Their preliminary architectural report is completed, and will inform the proposed solicitations.

1.  "Recommendations to the National Science Foundation from the Board of FARNET, Inc. Regarding Inter-midlevel Connectivity after the Expiration of the Current NSFNET Backbone Agreement", FARNET, Inc., Waltham, MA, 11/91

2.  "EDUCOM  Networking and Telecommunications Task Force Statement on the Structure of the National Research and Education Network", EDUCOM, Washington, DC, 10/91


Plan Overview

An analysis of the tasks performed by Merit and its subcontractor Advanced Networks and Systems (ANS) under the existing cooperative   agreement suggests a resolution of the stability/competition dilemma. In addition to furnishing and operating telecommunication circuits and packet  switches, Merit staff serve in a higher-order technical capacity known as the "Internet routing authority"  (the tactical and technical maintenance of  the database that drives the dynamic packet routing algorithms of the worldwide  Internet). Although Merit now carries out both functions, the NEG have pointed out that it is not necessary they be vested in the same organization; the DAC observed  that in the case of multiple awardees for connectivity, separation of the routing authority function is desirable in order not to give one connectivity awardee a tactical advantage over the other(s). Since the provision of circuits and switches is highly competitive, but the key to network stability lies in careful and conservative operation of the routing authority, the NSF will address the issue of stability vs. competition post-1992 by issuing two solicitations: one (for connectivity)  crafted to promote competition, and a second (for the routing authority) designed to maintain continuity and stability. These solicitations will be developed with community consultation and advice, and the resulting proposed awards brought to the National Science Board for approval.

Over the past five or more years, the telecommunications industry has been developing a new set of standards for digital communication; these standards extend to speeds of 2.4 gigabits per second, and their adoption and implementation are likely to fundamentally alter the ways in which computer communication is done. Vendors have begun implementing the standards in switching equipment, and early examples are being installed by the carriers. The awards to be made under this Plan will be structured as Cooperative Agreements so that these new technologies, such as the  Switched Multimegabit Data Stream, (SMDS), Frame Relay, and others can be experimentally incorporated.

There are several ways to foster competition in Internet connectivity. NSF staff intend to follow the recommendation of their advisors by specifying, in the competition for Backbone connectivity services, that "more than one award will be made.

Since the major telecommunication carriers have begun to move aggressively into the Internet arena, effective and sustained competition is likely.


3 Feb 92     draft  solicitations  for  connectivity  and  for routing authority prepared
4 May 92    solicitations finalized, mailed out
20 May 92   public information meeting for proposers
3 Aug 92     proposals due
10 Sep 92    merit review panels meet
Oct 92        as required: site visits, reconvene panel
Feb 92        present recommendations to National Science Board
Apr 93        awards made

Cost Projections
                    Activity        Year 1   Year 2   Year 3
                                           ($m)     ($m)     ($m)

          CONNECTIVITY    $6.0     $5.0     $4.0
ROUTING AUTHORITY     1.2      0.9      0.9

These projections anticipate a decreasing schedule of costs for high-bandwidth services from the telecommunications carriers, and allow for equipment acquisition by the Routing Authority in its first year of operation.