EVALUATION OF TWO STATISTICAL CAC METHODS FOR VARIABLE BIT RATE TRAFFIC SOURCES

The aim of this paper is to present a complex comparison of two methods of connection admission control in ATM network. In the introduction the value of tra‐c management is brie∞y presented. In the next part, selected CAC methods, which have been compared, are presented. The following part is dealing with comparison of selected CAC methods on the basis of simulations in MATLAB for difierent sources of variable bit rate (VBR) tra‐c. K e y w o r d s: ATM, tra‐c management, connection admission control methods ATM (Asynchronous Transfer Mode) is a powerful and ∞exible switching and multiplexing broadband technology that allows concurrent transmission of voice, video, pictures and data. It is based on fast cell switching with a constant length of 53 bytes. The ability to deflne the provided Quality of Service (QoS) is one of the main advantages of ATM technology. This capability is implemented on the highest level in comparison with other broadband networks. Hence, the customer can choose the required QoS level and pay only for what he or she really demands. Tra‐c control mechanism | tra‐c management | should be implemented in the network to secure QoS. The ATM technology ofiers several tra‐c control mechanisms [1,2]. Tra‐c control can be based either on preventive or on reactive functions. Preventive trafflc control is mainly used in ATM networks and it utilizes functions as eg Connection Admission Control (CAC), Tra‐c Shaping and Network/User Parameter Control (NPC/UPC). A large number of various CAC methods have been ofiered but none of them is universal for all tra‐c types. Considering the fact that ATM is a connection oriented mode, it is possible to secure QoS for connections on the level agreed in advance. The required QoS level is arranged in the connection set-up through tra‐c contract. The quality of Service means provision and observation of required tra‐c parameters (cell losses, cell delay, etc) in the course of connection. The tra‐c contract states duties for both parties | network contracting to support the tra‐c at the arranged level and user’s agreement not to exceed the arranged performance boundaries. Connection Admission Control CAC is a preventive function of tra‐c control that determines whether new incoming connection to the network will be accepted or rejected. The decision process is based on various tra‐c parameters, such as the current tra‐c load, the values of characteristic parameters (eg mean and peak rates), availability of network resources and required QoS of existing and new connections. Many CAC methods and their modiflcations have been ofiered. Some of them are based on mathematical models, statistics and theory of probability. Another large group of methods is based on on-line tra‐c measurements and on the analysis of the bufier states in the network nodes [3]. Methods based on artiflcial intelligence eg fuzzy systems [4] and artiflcial neural networks are also developed and extensively studied.