20 January 1994 Interim neutral interconnect structure in Europe ------------------------------------------------ A joint statement by Howard Davies, Dante Frode Greisen, Ebone Glenn Kowack, EUnet Christian Michau, Renater Bernhard Stockmann, KTH Peter Villemoes, NORDUnet 1. Introduction --------------- With the growth of the Internet and with the growing number of service providers it is mandatory to develop the interconnect structure. The pilot GIX (Global Internet eXchange) in Washington (1,2) has gone in production in 1993 and new plans for a D-GIX (Distributed GIX) structure are being developed (3). It is expected that the D-GIXes will use the same routing registry data base as the GIX, but that tools will be developed which will allow the derivation of different routing databases for the D-GIX route servers to describe the situation at any particular D-GIX. However, as this is development and pilot work, there is a need to agree on interim solutions which support the overall strategy towards improved interconnectivity. The authors of this paper met to discuss how the service providers in Europe could best take advantage of the D-GIX development and also to set up guiding principles for the providers on how to cooperate until the D-GIX technology is fully developed. Regardless of the D-GIX development, a number of interconnecting points exist, eg. ethernets where different providers connect routers to exchange IP traffic. Such points are sometimes called DMZs (De-Militarized Zones) or neutral interconnects. We suggest a new name for such interconnection points that satisfy certain criteria: INIX (Interim Neutral Internet eXchange). For definition of other terms, such as GIX, route server, D-GIX, peering etc. see reference 1,2 and 3. 2. INIX definition ------------------ An INIX is an IP network connection structure that satisfies the following three conditions: 1. Freedom of peering. Each partner may bilaterally or multi-laterally decide to peer (send traffic to and accept traffic from) or not to peer with any other partner connected to the INIX. The partners are also free to establish transit agreements or not. 2. Potential GIX access. In order to minimise the undue proliferation of INIX sites, an interconnect point must demonstrate at least one access possibility to the GIX in Washington in order to qualify as an INIX. The organization owning the INIX may have such a connection itself, or one of the network connected to the INIX may have it so that a transit access can be offered to other networks, on terms to be determined by the network having the connection. 3. Openness and fairness. The owner of the INIX must offer any service provider the possibility to place a router at the INIX. This may be at a cost but such cost must be the same for all networks. Different INIXs may have different costs and tariffs and these tariffs as well as eg. the geographic location will determine whether any particular INIX will attract connections or not. INIXs may not make access conditional on the purchase of further services. Note that an INIX is not necessarily a D-GIX as a route server is not needed to exchange routing and payload traffic and thus not a condition. However, we expect that INIXs develop into D-GIXes, see below. 3. Strategy for European IP providers. -------------------------------------- At present, three announcements have been made in Europe of interconnects which seem to fit our INIX definition: - NORDUnet at KTH in Stockholm (contact Bjorn Eriksen <ber@sunet.se>) - Renater in Paris (contact Christian Michau <michau@urec.fr>) - Surfnet in Amsterdam (contact Erik-Jan Bos <erik-jan.bos@surfnet.nl>) The authors recommend that service providers use these INIXs to establish peering where improved connectivity is required. Note that the GIX is an INIX according to the above definition but that our recommendation is that the European INIXs are used to exchange production traffic between European service providers. We further recommend support of the D-GIX development work in agreement with the overall strategy of improved, cost effective connectivity. Finally, we recommend that the INIXs develop into D-GIXes, or IXFs or whatever will be the name of the structure which will provide optimal routing stability and connectivity support for Internet service providers. This is because service providers will need to invest in lines and equipment to connect to the INIXs and such investments will only be justified if they can be used over a long period. We have no recommendation regarding the number of INIXs. On the one hand, a large number of INIXs may minimize line costs if suitable transit agreements develop, on the other hand, the D-GIX development and eventual operation may become more difficult if there are too many locations in the structure. We could not identify an organization that would formally decide where INIX should or should not be located, rather we expect that market forces will determine the number and location of the INIXs. (1) Almes, Ford and Lothberg: Proposal for Global Internet Connectivity, 1992. (2) Bates, Karrenberg, Lothberg, Stockmann and Terpstra: Internet Routing in a Multi Provider, Multi Path Open Environment, 1993. (3) Eidnes: Draft Distributed GIX Specification, 1993.
participants (1)
-
Frode Greisen