High Speed 34 Megabit Atlantic Peering BOF (PEERING) Reported by Howard Davies/DANTE Introduction and Background The BOF was led by Howard Davies. Peak attendance was about 25 people. Howard explained that the origin of the meeting was an e-mail message sent about two weeks before to the US members of the CCIRN suggesting that the IETF would offer a good opportunity to initiate a discussion on the organisational and funding issues surrounding the establishment of high speed (34/45 Mbps) links between Europe and the US. Background material extracted from the original message was issued to the attendees and is annexed to these minutes. The main agenda was then a series of questions (given as headings below) which were discussed in turn. What Is The Demand? User demand is difficult to estimate. All the existing transatlantic links are heavily loaded. When DANTE introduced an additional T1 circuit in November, it was running at 70% load within 24 hours; demanding applications see poor service levels. All these factors point to unsatisfied demand. Total transatlantic capacity is currently 18 Mbps with the addition of a further 9 Mbps committed during 1995. Organic growth will quickly take the total requirement into the 34/45 Mbps range. It is not clear whether the introduction of new (multi-media) services will generate an additional step increase or whether they will just add to the existing organic growth. Trialling and operation of such services will need committed bandwidth if they are to be useful and not swamped by general purpose traffic. What Justification Is Needed For Funding? The US agencies have stressed the need for services to be seen to support mission objectives; quality of service is therefore an important issue. Funding can only be provided for things that are of direct and visible benefit to particular programmes. The individual requirements of each funding agency need to be clearly understood when formulating proposals for new or enhanced facilities. 34 or 45 Mbps? There appear to be no technical problems in using either 34 Mbps (European standard) or 45 Mbps (North American standard) circuits for connecting the two continents. The choice will therefore be determined by cost/price considerations. Cost Sharing Procedures (Via Service Provision?) No-one expressed objections to the principle of purchasing services (and even shared services) as a way of optimising use of available resources as long as certain important conditions were met; namely that the service should be clearly specified, that service levels should be guaranteed, and that there was confirmation that the service had been delivered. Given the management limitations inherent in IP, it is difficult (if not impossible) to meet all these conditions in an open, shared IP network service. On the European side, the reluctance of the European Commission to fund infrastructure provision might be turned to advantage by expressing all funding proposals in terms of project support. Scope For Collaboration With NSPs? Although there may in principle be some scope for getting favourable treatment from the NSPs, for example for relatively short periods after new capacity comes on stream and before it is taken up through normal commercial sales, the value is limited both in time and to the small segment of the research community which is prepared to trade stability for the opportunity of using high bandwidth. Most members of the research community, in contrast, give priority to stable services with good quality of service which can only be provided on reliable infrastructure. The NSPs have no interest in providing such infrastructure services on anything other than normal commercial terms. There is an NSP view, however, that the research community is asking for high speed services at zero charge; offers of payment at realistic levels may provide the starting point for successful negotiations. US Procurement Rules The fear appears to be unfounded that US and European procurement rules may each impose (conflicting) constraints that make it impossible to achieve a common decision on an interconnection service. Although both systems insist on demonstrating that good value for money is being obtained, usually by competitive tendering, there is enough flexibility to allow a procurement involving suppliers in both continents to conclude successfully. Further Discussion Following the completion of the main agenda, there was a wide ranging discussion in which the following points were made (among many others): o Europe will be at a disadvantage to the US in relation to reaching agreement on equitable cost sharing as long as the only global exchange points are in the US. o High line prices in Europe continue to be a major distorting factor in planning rational network configurations; in particular, the cost of distributing traffic within Europe eliminates any advantage of using cost effective, high capacity transatlantic links to carry traffic to more than one country. o Until the line price problem in (b) is resolved, (a) will continue to apply. Conclusion The principal issues identified were: o The importance of quality of service in specifying requirements and in formulating proposals to meet them, o the need to understand in detail the specific requirements of the different funding agencies and the groups they support, and o the value of justifying proposals in terms of project/mission support on the European side as well as in the US. It was agreed that these issues should continue to be studied and should be topics for further discussion at the next IETF and the June 1995 meeting of the CCIRN. Annex: Background Information An update on the current European situation is as follows: 1. There are a number of activities which have the common aim of promoting the establishment of 34 Mbps and higher speed services for the research community in Europe. The main activity is the EuroCAIRN Project on which nearly all Central and Western European governments are represented. 2. In July 1994, DANTE was contracted by EuroCAIRN to perform a study which included surveys of user requirement, the state of the technology in relation to operational services, and likely availability of services from the PNOs as well as the development of an implementation plan. 3. As a result of our work on this study, we are now in a position to proceed to the implementation of a (geographically) limited service in collaboration with a group of national research networks that are in a position to go ahead quickly. 4. The principal obstacle to progress is now the lack of availability of 34 Mbps international circuits and services and uncertainty about the price levels that the PNOs will have in mind when they do deliver them but we are optimistic that, with political support via EuroCAIRN and other channels, progress will shortly be made towards the start of a service during 1995. 5. An important conclusion from our survey of European research networks for EuroCAIRN was that high speed connectivity to North America is seen as a vital component of any pan-European service. We therefore need to plan how this connectivity will be provided in parallel with our other activities. For resilience reasons, we see the provision of two independent links across the Atlantic as an initial minimum. 6. Our preference, both within Europe and between Europe and North America, is to purchase services from commercial providers though we accept that in the short term some form of collaboration in which we take some responsibility for technical and operational networking arrangements within Europe may be necessary. 7. DANTE is quite prepared to work within the kind of commercial framework adopted by NSF. We already have contracts with ANS and Sprint for the management of the US ends of our 2xT1 and 1xE1 transatlantic lines and we are also engaged in discussions with MCI on the provision of further T1 capacity. The discussions include an exploration of the possible ways of organising joint funding. 8. A big advantage of moving away from the concept of funding bodies supporting particular lines is that there can be more effective routing of traffic between US and European locations independently of who is ultimately funding which traffic. In order to plan high speed connectivity across the Atlantic, we need to understand how the US funding bodies wish to approach these issues. Some form of cost sharing will be essential as a fair way of distributing the actual costs between the parties that benefit. We also need to coordinate our approach to the commercial service providers.