The Norwegian research network
  
  In Norwegian Search:

Supernett ATM

This project started in 1995 as part of a co-operation agreement with Telenor. During 1995, ATM-technology was chosen and equipment was procured. Implementation started in November 1995. The changeover from phase 1 (with IP-routers) was gradual, starting with the Trondheim-Oslo segment ending with Trondheim-Tromsø in February 1996.

The acceptable-use policy for Supernett ATM has been relaxed so that reasearch projects in commercial companies can take part.

UNINETT offers support for researchers and students at Universitites and Research Institutes that want to experiment with Supernett. No special project approval is necessary, and UNINETT offers financial support for interesting proposals.

The participating orgaizations are the Universities of Tromsø, Trondheim, Bergen and Oslo, SINTEF, The Norwegian Computing Centre(NCC), The National Library in Mo I Rana and Telenor Research.

The network

Supernett, with its customers, encompasses in total about 17 ATM switches. Around these, there are 9 routers and about 40-50 workstations with ATM interfaces.

A sketch of the main parts of the network is shown at http://www.uninett.no/info/nett/supernett/sII-no-kart.gif

The backbone of the network is based on 34 Mbps chunks from spare 140Mbps systems at Telenor. In case of a severe outage in production transmission systems, Supernett connections could be taken down to serve as a backup. This force us to provide additional leased 2 Megabit/s lines for our own backup. On these lines, we wanted to run our non-university traffic like that from colleges and research institutes.

This turned out to be a complex routing task and we had to develop a set of linear equations solved by a mathematics program to satisfy the routing policy imposed on us by acceptable use and network engineering.

The IP trafikk to and from the Universities has a dedicated capacity of 15 Mbps. The rest of the capacity is free to be used by ATM experiments.

Background

Supernett phase 1 from 1992 was based on 34 Mbps connectionsusing IP-routers at PTT-locations for the backbone between University cities and FDDI rings to feed campuses. It was a successful demonstation of what was at that time high-speed communication.

In phase II, we investigated the ATM technology because it offers a more flexible bandwitdth allocation and because it can integrate data traffic and telephony in the same network connections. ATM also easily scales to higher bandwidths than are used today. ATM can run multiple parallel virtual networks on the same backbone and the capacity of each can be dynamically adjusted by software. The necessary capacity can be negotiated by signalling from the customer equipment to the network operator. ATM technology can therefore provide a practical and economical use of available transmission capacity.

The idea behind ATM is that the smallest transmission unit is a cell, which is 53 bytes, and thus small enough to accomodate traditional telephony data that need low transmission delays and variations in delays to give good sound quality. Every cell is marked as belonging to a connection, thus the service can be adjusted for different types of data.

IP packets are transmitted by chopping them up into cells, and this wrapping in cells has a cost in lost bandwidth. The overhead as measured in our network is as much as about 20-30%, depending on the type of traffic.

The integration savings need to be better than that to make ATM favourable, seen from the IP side.

Applications

New applications for Supernett are a distributed national filesystem for supercomputing, projects in the HUGIN-programme like video servers (LAVA) and computer supported cooperation tools (LAVA). LAVA factilitates quick search, navigation and replay of film material over the network. Participants are the Norwegian Broadcasting Coorporation, The Norwegian Computing Centre and SINTEF. BATMAN is investigating various cooperative work support tools like videoconferencing and shared workspaces.

From Supernett phase 1, we have IP traffic between universities, classroom videconferencing, multimedia multicasting (MBONE), and high speed interconnection of and access to supercomputers in Oslo, Bergen and Trondheim. The added capacity of Supernett has been of special importance to improving the weather forecast computations by DNMI by enabling them to increase the resolution of their datamodel.

In parallel, we have done experiments on the network level with ATM-like signalling and performance measurements. We used workstations with ATM-cards, owned partly by UNINETT and partly by universities. We are tunnelling ATM signalling for the dynamic setup of connections (SVC). Telenor is not providing this service yet in the core network. This work was presented at the UNINETT-96 conference.

Our experiences is that ATM is far from being fully ready for production service. Both standards and products are still short of fully supporting the ideas og dynamic assignment of bandwidth and easy end-to-end connection setup.

Internationally

Supernet ATM is connected to JAMES, the European ATM pilot backbone constructed by Telecom operators. Through JAMES researchers at universities are able to participate in experiments for projects for the 4th Framework Programme for the European Union. Among them are AQUARIUS( NTNU and University of Bergen), MERCI (University of Oslo) and TRUMPET ( Norwegian Computing Centre).

UNINETT has participated in TF-TEN, a sub-project of TEN-34. In that project we are exploring the basic beahviour of ATM and how IP behaves over ATM.

More info about Supernett can be found at http://www.uninett.no/info/nett/supernett/.


info@uninett.no 2004-01-22