Last update at http://inet.nttam.com : Mon May 22 12:31:28 1995 SuperJANET and its Real User Applications John Dyer Abstract The SuperJANET multi-service broadband network has been in place for over two years. During this time real users have been able to take advantage of its advanced capabilities, bringing WAN like performance to the national network. This paper reports on its architecture, operations and usage. 1 Introduction SuperJANET is the UK Academic Community broadband multi-service network. It is comprised of an ATM element running at speeds up to STM-1 (155 Mpbs) and an SMDS element running over 34 Mbps fibre optic links. The SuperJANET network reaches over 50 sites and is connected to the existing JANET network bringing the total number of sites connected to directly to the UK Academic Network to well over 240. 2 Network Installation The original JANET network used to run an X.25 connection oriented service running first at 64 kpbs and later at 2Mbps and 8Mbps in the core. As demand grew to support more traffic intensive applications it became apparent that a broadband network would be required to support users. A case was developed and supported by the funding body which made money available for a four year project (now in its third year). A decision was also made to mount an IP service (JIPS) over the wide area network around the same time due to the rise in popularity of applications utilising the IP protocol suite. Tenders were invited for the installation of the new network and a contract including collaborative exploration was awarded to British Telecom (BT). The installation took place in four phases each of which are described briefly in the following paragraphs. 2.1 The Pilot Phase The Pilot Phase of SuperJANET during the 1st quarter of 1993 involved the installation of 140 Mbps PDH lines to six sites. BT provided hardware multiplexing at both ends of the lines hence delivering four 34 Mbps tributaries between each site and the major BT switching node at Birmingham. This arrangement, although quite basic has been very useful enabling the team to reconfigure the PDH network topology at will. 2.2 Phase One By September of 1993 another seven 140 Mbps PDH sites had been added making thirteen in all. Additionally SMDS with an access class of 10 Mbps was provide in parallel to all of these sites. 2.3 Phase Two During 1994 the network was further expanded by the installation of around 40 further SMDS sites each with a 10 Mbps access class and the installation of SDH (STM-1 155Mbps) lines to five of the PDH sites in parallel with the existing multiplexed 140 Mbps lines. 2.4 Current Phase During May to July of 1995 SDH is being installed to all remaining PDH sites, some sites being allocated a single STM-1 line, others getting dual STM-1 connections providing them with an aggregate bandwidth in excess of 300 Mbps. It is expected that the existing PDH 140 Mpbs lines will remain in place in parallel with the SDH for some time, providing additional bandwidth and resilience in the early days of experimentation. 2.5 Equipment The PDH lines have been fed into GDC DV2 switches by the use of E3 PDH interface adapters. Additionally 100 Mbps TAXI and 2Mbps E1 interfaces have also been provided for the connection of end stations, routers and codecs The first five SDH sites are currently operating on FORE ASX-200 switches connected by monomode interfaces to the WAN and multimode interface to the local area. Interconnect between the SDH and PDH is provided by the connection of DV2 and ASX200 TAXI interfaces. 3 Network Usage The usage of the network falls into one of the following main categories : * Constant bit rate (cbr) video traffic. * IP applications traffic generating variable bit rate (vbr) traffic for transport over the ATM network. * Direct ATM application traffic and ATM research traffic. 3.1 Video traffic One of the major objectives of SuperJANET was to permit the transmission of high quality full screen real-time video signals with accompanying audio. At the time of inception of the service, the best way of providing this service was considered to be by the use of dedicated hardware codecs. These codecs are usually associated with dedicated confrencing rooms, video suites or other purpose adapted rooms. The codecs are connected to normal analogue audio and video equipment which encodes/decodes H.261 data streams which are transmitted across the network by means of emulated 2 Mbps circuits. The video streams are delivered to one of three video switches (MCUs) which are used to control the set- up and management of the conferences. At the moment each codec is allocated a permanent virtual circuit (pvc) to a particular MCU port on a fixed basis, however when switched virtual circuits (svc) become available more effective use of the video resources will be possible (particularly the MCUs). The MCUs can provide a sound activated switching capability which is used to select the video signal being broadcast to the other participating sites. 3.2 IP applications running over ATM The requirement to support traditional IP applications from workstations is essential as most user network traffic is generated from IP based application programmes. I feel it is very unlikely that we will see an overnight migration of all IP traffic to become direct ATM traffic. It is far more likely that traditional IP traffic will remain for some years to come and be carried encapsulated over the ATM networks. Initial experimentation had begun at the date of the writing of this paper. Six national sites have installed routers connected to ATM switches passing pilot IP traffic over the ATM WAN. By the time of the INET conference, it is confidently expected that a portion of the JANET IP production traffic will be traversing the ATM WAN as a matter of course. Additional sites will also be brought onto the service. 3.3 ATM Research uses of the network Apart from networking specialist experimenting with network specific applications such as network management software, little direct end user use of the ATM functions of the network have yet been put into every day use. This will change as appropriate product becomes available - In particular I expect this to happen first with devices like vbr codecs with built in adaptation layer support which are just reaching the market place in the UK. I feel confident that other application areas will follow 3.4 Problems, Problems... The only major technical problems that we have suffered relate to the concurrent transmission of cbr video data and vbr IP data on the same physical interfaces. Initial experiments with 2 Mbps cbr video data had gone well and provided satisfactory audio /visual results until the routers were allowed to introduce bursty vbr data from computer generated sources. The effect of this was to produce intermittent loss of synchronisation of the hardware codecs. The source of this break up was eventually traced to the ATM switches which had only been provided with single buffers per interface. The effect of this was that when cbr and vbr were mixed and the vbr produced congestion such that some data was discarded, both cbr and vbr was discarded together. The loss in cbr video data causing the loss of synchronisation information. Re-synchronisation was usually re-established by the codecs themselves after 20-30 seconds. The switch manufacturer was helpful in sorting out the problem and eventually installed dual queue interfaces in all our switches thus removing the problem. The system has been working with concurrent cbr and vbr on the same physical interfaces ever since the interfaces were replaced although extreme testing still remains to be undertaken. The cbr video is allocated to a high priority queue and is preserved, the vbr computer data being discarded in cases of congestion. A second problem that continues to bug us is the ineffective or non- existent traffic shaping (and re-shaping) that we would like to see in our switches. If traffic remains unshaped and cells arrive at a switch in bunches (which often happens after traffic aggregates after passing through several different switches), unnecessary traffic congestion and discarding can take place particularly when vigorous policing of the data is carried out. In cases where the data is policed hard, it is essential to have good traffic shaping facilities. This is not yet provided in all equipment. 4. Usage of the Network SuperJANET provides an advanced resource for the researchers of the UK. It is being used in a number of national application programmes, however it is also used as an access path to the European wide ATM trial that has been put in place with the collaboration of the European Telecom Operators. 4.1 Application Areas As previously reported at INET, SuperJANET has been used to support applications in the areas of * Distance Learning * Electronic Publishing * Remote Access & Participation * Audio-Visual Communications 4.2 Teaching Surgery Major progress since the first report is the complete acceptance of the use of the national full screen video services to support the teaching of surgery. Back in 1993 we ran a pilot with a few pioneering clinicians. Now there are regularly timetabled and structured courses which make extensive use of the networked teaching facility and its use has been written into the ciriculuum of every major teaching hospital on the network. Lectures and examples of operations are undertaken or prepared by the sites with specific specialisation and broadcast over the SuperJANET network during every teaching week of the year. Surgeons are able to transmit a perfect view of the operation from their operating theatre using high quality cameras situated over the operating table. The view obtained far exceeds what a junior medical student can expect to see from the back of the theatre. It also reduces the risk of infection of the patient. The data stream is transmitted over the SuperJANET network where it is viewed by students at the multiple remotes sites. The students at each site are able to ask questions of the clinical team making the broadcast questions during the procedure by the use of return audio paths and in some instances also have camera control. This interactive element provides an essential advantage over what could be provided by a linear video taped lecture. The clinical tutor is also able to ask for supporting material to be mixed in by the use of a laserdics player. One of our sites also archives the teaching material to tape for later use for students unable to attend the live interactive transmission. 4.3 The European ATM Connection The European PTTs have co-operated to put in place a short term technical trial of ATM across national boundaries. The network sometimes known as the MoU involved initially twelve countries and continues to expand. The objective is to simulate real user traffic by having real users do live experiments across the network. An example project is VISINET in which UKERNA is a partner. VISINET is a virtual prototyping application designed to allow virtual reality applications to be used between partners to shorten the industrial design process. The one year project involves contributors from The Netherlands, United Kingdom, Belgium and Portugal. The workstations used in the project have been connected to the national research networks using ATM equipment and then internationally by the MoU ATM network. 5. Conclusions ATM is still in its infancy. If it is to be driven forward so that the users may make full advantage of the multi-service facilities promised by ATM, then forward looking funding bodies must invest in the technology ahead of the commercial game. Only in doing this will researchers be given an edge that can enable them to keep ahead of the field in their chosen area of research. The message is that no one yet understands all the problems and even less is able to offer product that solves them, but progress is being made. Even the embryonic superhighways of today have a lot to offer the user. Acknowledgements Grateful thanks is given to the many members of the UK Academic Community who have given so much time and expert effort to work on the SuperJANET project and have made it a success. Thanks also to the past and present members of the UKERNA team and to my wife Sue who continues to provide the support that I canŐt get from AAL5. Author Information John DYER is the UKERNA Technology Manager and head of the Technology Group; a band of dedicated workers whose goal is to provide innovative leadership to the UK Academic Community networking programme. He studied at Oxford Polytechnic (now Oxford Brookes University) and Green College Oxford, a Post Graduate Medical College of Oxford University. He has been involved in academic networking both in the UK and Europe through JNT, RARE, TERENA and UKERNA for the past 10 years. You may contact John Dyer at the following address : John Dyer Technology Manager UKERNA The Atlas Centre Chilton, DIDCOT OX11 0QS United Kingdom Tel : +44 1235 822200 (after 15 June 1995) Fax: +44 1235 822399 (after 15 June 1995) Email : John.Dyer@ukerna.ac.uk