A Behind-The-Scene Look at NSFNET Support

By Denise Miller
IBM

Since the Merit Computer Network assumed management of the NSFNET backbone in July 1988, the network has undergone a number of significant changes. Among these are:

• Network redesign to increase network redundancy by providing multiple T1 circuits to each node
• Extended SuperFrame (ESF), a qualitative measurement on the physical lines of the T1 network
• Improved Ethernet performance (3COM)
• NNStat, software which provides a flexible method for gathering traffic statistics, querying interactively, and remotely collecting them for later analysis
• 4Mb memory expansion in the routing control processors

These enhancements to NSFNET illustrate the extensive planning and cooperation between the three-way partnership of Merit, MCI, and IBM.

IBM's role

From IBM's standpoint, more is involved than just providing hardware. Mike Sabol, who coordinated the IBM team's effort, works in the University and College Systems group and is located in Gaithersburg, MD.

We asked Mike what he feels is the most challenging aspect of an upgrade effort. He responded with a case in point, the 4Mb memory expansion upgrade:

Because of the astronomical growth of the network, we needed more memory installed in each RCP. The Route Control Processor allows the NSS (Nodal Switching Subsystem) and the rest of the NSFNET to adapt to changes in topology due to line problems, machine failures, or changes in traffic patterns. I look at the RCP as the heart of the NSS. How do you install this up-grade without the NSS node being taken out of service for an extended period of time? To plan and make the change with minimum downtime to the network is the challenge.

The process of installing an upgrade

What is the process for effecting such changes? Sabol continues:

Before any up-grade or change is made to the NSFNET production network, it is first installed and fully tested on the Research and Test Network. After the change is approved, it is then moved to the production network.

In the case of the memory expansion, IBM worked closely with Merit to select off-peak hours and notify the local site personnel and IBM National Service Division Customer Engineer's (CE). The local IBM CE actually performs the upgrade with support provided by Merit Network Operations Center (NOC) personnel and/or myself. The CE's are the ones that make my job easy.

Marti Arsenault, a host site field manager, is Mike's interface to the IBM National Service Division. Marti described the role of host site field manager:

The host site field manager has responsibility for the nodes in the network and provides a single point of contact to the NOC. We also work closely with the NOC to ensure that all deadlines have been met.

This involves an Ann Arbor CE, Jerry Breuer, dedicated to the NSFNET project during installs and day-to-day repair actions. Jerry performs a crucial part in this activity, dealing with the remote IBM field managers and CE's, interfacing with Merit, MCI and other IBM organizations which helps to make this operation run smoothly.

According to Sabol, it takes teamwork to avoid network downtime:

"A good example would be the T1 reconfiguration. That change required a total effort by Merit, MCI and IBM. We literally rebuilt the entire network and at the same time kept it running.

IBM CE's installed all the hardware that was required at the NSS nodes, at the same time that MCI technicians were testing the new circuits. During circuit cut-over, IBM CE's were at the sites on standby to support Merit and MCI with the cut. That's what I mean by teamwork."

The Merit NOC

Elise Gerich of Merit/NSFNET commented on Merit's role in the network changes:

The Network Operations Center oversees all changes to the NSFNET backbone. Since the NOC constantly monitors the status of the backbone, it is ultimately responsible for insuring the smooth implementation of any changes.

Because of the close cooperation and good communication between IBM, MCI, the mid-level NOCs, and Merit's NOC, the NSFNET project team has been very successful in maintaining better than 99% up- time for the network as a whole.

MCI's role

MCI's role was explained by Russ Shorter, MCI Senior Engineer:

"We provide the backbone circuits, design the network physical topology, and participate in all research and technologies relative to the physical network. We also provide physical network management capabilities and features to the NOC for the backbone network. Our staff performs installation, maintenance and support during enhancement activities."

Planning for the unexpected

What happens when things don't all go as planned? Mike Sabol commented: "You have a backup plan in place for the unknown. But then, how does one plan for an ice storm in Houston, or when an IBM CE is trapped in his car by a skunk at 5 a.m. in the morning?"

Regarding future enhancements, Sabol commented: "Around the first or second quarter of 1990, we are planning on installing Cylink 4201 Advanced Channel Service Units which at the present time, are installed and being tested on the NSFNET Research and Test Network."

As you can see, making enhancements to the network in such a way that minimum or no down time is felt by the end users takes tremendous cooperation and coordination between the staffs of the partnership.

Indeed, there are even more people involved behind the everyday operation. This combined "people-power" ensures that the NSFNET backbone will continue to be a viable, state-of-the-art, network, responsive to the changing needs of its users.

Taken from The Link Letter, Vol. 2 No. 7, January 1990.