No abstract

A key objective of ATM-based networks is to provide at the same time guaranteed QoS to real time and non-real time services. This calls for thoroughly engineered traffic control methods for those service categories as well as for the overall integration strategy. The main objective of this paper is to investigate CAC in an ATM testbed with real switches and as realistic traffic as possible. The used traffic is video (MPEG model based on traces) and data modelled as traditional on/off sources. The video traffic is given priority over the data traffic. To complement and verify the experiments, two simulation tools using as input the artificial MPEG models and the real traces have been developed. The CAC boundary for the nonpriority case has also been derived analytically. The experimental sources are modelled as discrete-time Markov sources. A matching method to avoid state space explosion for the superposition is applied on the source models.

Abstract. This paper deals with the experimental investigation of a preventive traffic control framework for ATM networks as implemented in the EXPLOIT Testbed. This framework comprises a Usage Parameter Control (UPC) function and a Connection Admission Control (CAC) function. The obtained measurement results are complemented and compared with simulation studies and analytical results. For several OdOff traffic types it is shown that the implemented UPC function enforces both the Peak Cell Rate (PCR) and the Sustainable Cell Rate (SCR) simultaneously. By measuring the cell loss ratio as a function of the number of multiplexed sources, admission boundaries for a given network performance objective can be found and compared with the implemented CAC function based on the convolution algorithm. It appears that network performance objectives can be guaranteed, while exploiting most of the possible multiplexing gain in a network with small buffers like the EXPLOIT Testbed. Finally, it is shown that the two co-operating functions constitute a traffic control framework which is robust against traffic contract violations as well as worst case traffic utilising the contracted UPC function tolerances maximally at both the cell level and the burst level.

No abstract