New England GigaPoP and High Bandwidth Network Project

Project Description

In this project, which is funded (Grant No. NCR-9617127) by the National Science Foundation under the Connections to the Internet program, Boston University will establish a framework to enable our institution and other institutions to acquire high-bandwidth connectivity among themselves, to the vBNS, and to the Internet. Our purpose in doing this is both to put in place the sort of high bit rate connectivity we need in order to conduct certain research in computational science and networking, and to begin constructing the infrastructure for the next generation network in our region. A substantial part of the research and educational activity we will use this high bandwidth network for, both within Boston University and in conjunction with other institutions, is associated with MARINER, our NSF-funded Metacenter Regional Alliances program. There are three components to this project:

High Bandwidth Metropolitan Network: NYNEX, the regional Bell operating company, plans to establish an ATM cloud in the Boston metropolitan area by mid-1997. This ATM cloud will operate initially at OC-3c level, and connections to it at that speed (and lower) will be available at an aggressively priced tariffed rate. This will enable us to establish a high-bandwidth metropolitan area network interconnecting local institutions that will allow our collaborators to make effective use of our facilities and generally enable the kinds of regional activities that are impracticable in the context of the current Internet. (For want of a better term, we will use the "metropolitan network" to refer to this area ATM cloud, even though we anticipate that it will grow beyond what is strictly the metropolitan Boston area.)
Metropolitan Link to the vBNS: In addition, we will bring the vBNS into Boston and into the metropolitan network at OC-3c level. We have our own institutional needs for high bandwidth connectivity to the national centers and to a number of associated researchers in various parts of the country (e.g., see representative list of research projects below). Besides our own needs, however, it is our intention to enable vBNS access for any institutions needing it that are connected to the metropolitan ATM network. We expect to apportion our cost of maintaining this connection across any such additional sites. Several models for apportionment are under consideration, from equal sharing of the cost to instituting a charge for bandwidth or total traffic.
Internet Connectivity via ATM: Once the metropolitan ATM network has proven stable, we will look for an Internet provider to make Internet access available over the ATM network. Our own current Internet connection uses a 10Mb Ethernet-over-microwave technology that we have had in place since we helped the vendor develop it in 1987. We are not far from loading it to its realistic capacity of about 4 Mb/s, so we expect to be needing a more robust pipe for this commodity traffic in the time-frame of this project. We are also aware of similar capacity problems encroaching on other institutions in the area. If we are successful in connecting a significant number of other institutions into this ATM network, then we should be able to leverage the needs of the whole group into getting favorably priced Internet service.

Campus Network Architecture

While the bulk of the University's internal network is 10 Mb/s Ethernet, we are already running both 155 Mb/s ATM and 800 Mb/s HiPPI networks for those applications that need higher bandwidth. The HiPPI network is currently confined to communication among the components of our 38-processor SGI POWER CHALLENGEarray and its associated graphics equipment, although it is also routed to the rest of the campus network. As the required technology becomes available, the HiPPI network will also be connected to our new SGI Origin2000 supercomputer. The ATM network, which will continue to grow over time, connects these high-performance facilities to equipment in the College of Engineering, the Departments of Chemistry and Physics, the Center for Remote Sensing, the Center for Neural Systems, and the Computer Graphics Laboratory (which also serves the Departments of Computer Science and Mathematics). Essentially any location on campus needing a high bandwidth connection can be readily served be extending the ATM network through our existing fiber plant. Two such extensions are already planned: one to the Department of Computer Science, which is about to install a substantial amount of ATM equipment in connection with an NSF Academic Research Infrastructure grant, and another to the Manufacturing Engineering/Fraunhofer Collaboration, which operates its own SGI Power Challenge in a computerized machining facility. Routing between the ATM network and the general campus network has been done through a conventional Bay router, but we are planning to use a NetStar router to provide high speed routing directly between the HiPPI network and the ATM network in the near future.

Metropolitan Network Topology

Our intention in establishing this metropolitan ATM network and vBNS connection is to plant the seed of the next generation of networking for the Boston area and as far beyond that as economics and the carriers will allow it to grow. Our expectation is that once it becomes immediately possible to connect to this network, and especially once we are at the point of opening it up for Internet access, it will begin to gather real momentum. We have seen this phenomenon twice before: with BITNET in the early 1980's, for which we created the original Boston hub; and with NEARnet in the late 1980's, which Boston University, Harvard and MIT put together initially for the handful of sites that were about to lose the ARPANET. The figure below shows in schematic form the evolution we expect the metropolitan network to go through over time. Initially (Stage 1), participating institutions will be interconnected via the metropolitan ATM network but will retain their existing Internet connections. Connections to the vBNS will be through Boston University by way of the metropolitan network. In the diagrams, we show application areas ("App") as gaining access through their campus network structures. While this is the most desirable arrangement and will be the case from the beginning at Boston University, we recognize that connectivity may develop differently at other institutions. It is possible that in some cases a research group or a department may be more prepared to make the first connection than an institution as a whole. Later (Stage 2), we expect to move Internet connectivity to the metropolitan network, with an Internet Service Provider providing high-speed connectivity directly to the ATM cloud. Finally (Stage 3), on the assumption that the vBNS will itself evolve into a national high-bandwidth research network, we would expect to take Boston University out of the path and interconnect the metropolitan ATM network and the vBNS directly.

Support for Quality of Service

We will actively work toward a rational approach for dynamically allocating bandwidth on the metropolitan network. In lieu of a well-established technology for controlling bandwidth availability and quality of service, our initial plan is to create an oversubscribed set of constant-, variable-, and undefined-bit-rate virtual paths interconnecting the various sites on the network and a corresponding web-based mechanism for "signing up" for various levels of use. We will work with NYNEX to move the level of control beyond the merely collegial by exploring the development of an administrative interface between their network structure and the participating institutions. It goes without saying that we will be constrained by the evolving ATM standards and the vendors' implementation of them in accomplishing this. While our expectation is that it will not be trivial to establish effective and efficient QoS mechanisms across a network that spans several administrative domains, ATM equipment from different vendors, and that will be used among a variety of organizations both locally and through the vBNS, we anticipate that emerging ATM technologies will address many of these issues. However it is done, it is essential that this network be made as dynamically and automatically manageable as possible. If the work on RSVP advances enough to perform this function and is appropriately extended into the ATM domain, we will of course adopt it. RSVP is also of interest to us internally in the context of providing some level of QoS guarantees for users that remain on Ethernet either at 10 or 100 Mb/s. Some of our researchers (Chlamtac, Crovella and Yates) have a direct interest in contributing to the development of RSVP.

Research Projects

This is a list of some of the research projects at Boston University that will use the high bandwidth network discussed here. Where applicable, the specific institutions with which high-bandwidth communication is required are also listed.

Traffic Controller Module for RSVP
Imrich Chlamtac, Electrical & Computer Engineering Department

Providing End-to-end Quality of Service in an IP-over-ATM Network
Mark Crovella and David Yates, Computer Science Department

Metacomputing and Distributed Visualization
Glenn Bresnahan, Scientific Computing and Visualization Group and Roscoe Giles, Electrical & Computer Engineering Department
NCSA and PACI partners, EVL

Earthquake Physics
William Klein, Physics Department and Roscoe Giles, Electrical & Computer Engineering
Cornell, IGPP-SIO (UCSD), Univ. of Colorado, MIT, Northeastern

Algorithms for Global Optimization, Enhanced Sampling and Reaction Pathway Determination
John Straub, Chemistry Department
NCSA, PSC

Modelling Excited State Chemical Dynamics
David Coker, Chemistry Department
NCSA

---------------------------

For further information about this project,
please contact John Porter (jporter@bu.edu)

---------------------------

Related Information

Boston University Home Page

Center for Computational Science

Scientific Computing &Visualization Group

MARINER Project