Abstract
In broadband multimedia satellite networks it is necessary to multiplex bursty streams of traffic with differing Quality of Service (QoS) requirements to maximise the utilisation of the satellite link bandwidth. Providing the desired QoS of each service, in a multi-service environment is a major challenge for satellite networks. Asynchronous Transfer Mode (ATM) which provides hard QoS guarantees is suitable for a multi-service satellite environment. ATM has been developed as a vehicle for multimedia communications and is widely regarded as one of the most important and fastest-growing communications technology of this decade. The design of suitable traffic and congestion control algorithms is one of the most important challenge for the success of an ATM-based satellite network. This thesis develops and optimises a traffic and congestion control mechanism which can provide users the required QoS for ATM over satellite networks. In order to provide QoS differentiation for end-to-end communication it is proposed to use both loss and delay priorities, which are determined form the required Cell Loss Rate (CLR) and Cell Transfer Delay (CTD) parameters, for each service class. A multiple shared buffer scheduling (MSBS) policy considering both delay and loss priorities, is proposed and evaluated for scheduling and discarding of ATM cells. It is shown that both the CTD and CLR requirements of ATM services can be met by the MSBS scheme. A combined preventive/reactive control scheme incorporating an adaptive Leaky Bucket (LB) is investigated for the satellite environment. It has been found that reactive control improves the cell loss due to congestion for time scales larger than the propagation delay. As the satellite air interface bandwidth is currently one of the most expensive commodities in the service provision, an adaptive MAC protocol that can support the ATM service classes whilst maximising the bandwidth utilisation, is proposed and evaluated. The mapping of ATM service classes to MAC classes and the use of a prioritised request queue provides the QoS differentiation required by ATM networks. It is shown that a pure reservation system performs poorly for very bursty user traffic. The user population which can be supported using Random Access (RA) for very bursty users with short burst duration is higher. The system throughput can be maximised, by making this protocol adaptive to changing traffic characteristics. It is shown that the utilisation of the frame capacity and the total number of users served can be improved by using this protocol.