Partner und Internationale Organisationen
(Englisch)
|
B, F, D, GR, I, N, PL, P, SI, CH, UK
|
Abstract
(Englisch)
|
Following the success of terrestrial mobile telephony, mobile satellite communications were predicted a glorious future in the beginning of the nineties. A number of research projects were initiated in the private, public and academic sectors to lay the technological foundations for such systems. The prospect was to provide GSM-style voice and data services to business travellers and to rapidly connect even the most remote areas on the planet. COST 252 was to provide input to standardization bodies and industry in the areas of systems engineering, protocol design and evaluation and radio link technologies. The leading idea of COST 252 was to select or design techniques that fit smoothly in the terrestrial infrastructure so as to enable integrated dual-system terminals and roaming between networks. The Telecommunications Laboratory of EPFL contributed a coverage analysis tool that takes into account terrain shape and can interface with common Geographic Information Systems (GIS). The aim was to test the promise of 'service anywhere-anytime'. For Low-Earth Orbit (LEO) systems, the time-statistics of satellite visibility led to the conclusion that real-time conversations cannot be provided with sufficient quality in heavily built-up areas as well as in hilly terrain. As time progressed, it became apparent that in Mobile Satellite Systems, the 'large systems' approach adopted by Iridium and others may not be economically viable. In fact, systems designed in this way need to launch a large number of satellites before they can become operational. Once in place, their capacity remains essentially fixed. The financial problems of all three 'large' Mobile Satellite Systems that have already launched satellites testify to the need for a more gradual roll-out. Alternative systems that seem more likely to succeed include the Indonesian 'ACeS' (Asian Cellular Satellite Communications System) built by Lockheed and the Inmarsat series of satellites. Unlike the troubled ventures mentioned before, these two systems use satellites in geostationary orbit (GEO), permitting market entry with but a single operational satellite. All along its involvment with COST 252, the Telecommunications Laboratory proposed several ideas on how to reduce economic risk. Chief among them is a protocol architecture for using bent-pipe (i.e. simple transponder-type) satellites instead of a satellite mesh topology with inter-satellite links (ISLs). In fact, all mobility-management functionality may be placed on the ground, effectively enabling this kind of architecture. But even when ISLs need to be used for other reasons, on-board routing can be made more robust against the rapid change of applications that is witnessed in today's networks. We propose to use a routing abstraction layer (RAL) in a packet switching network to reach maximum service flexibility. RAL can make the journey across many satellites look like one hop. It works by using mobility management for the satellite network independently of end-to-end mobility Towards the end of COST 252, there was growing interest in TCP/IP due to a number of factors. Beside the fact that it will probably carry the majority of traffic very soon, packet-switching comes naturally to satellite systems engineers since a form of packetization is needed anyway in today's TDMA channel access and radio link protocols. Finally, after the success of Hughes DirecPC and a research effort by the French to push Skybridge, providing Internet access at reasonably low prices seems to be possible. We therefore evaluated the performance of TCP/IP in the context of a system able to provide a low-economic risk entry into the Internet access market. Skybridge was chosen as a reference model, but results were later found to hold for most LEO satellite proposals, including the original Teledesic constellations. We show that for adequate buffering and a Bit Error Rate smaller than 10-7, almost 100% link utilisation can be attained. We also investigate effective throughput, a measure indicating how long it takes to transfer a given file, depending on the size of this file.
|
