ATM traffic control at burst level
Ilyoung Chong · 1993
BISDN (Broadband ISDN) is to provide a common user-network interface to various traffic such as voice, data, video. Informations generated from BISDN traffic sources are transported across a network. ATM (Asynchronous Transfer Mode) is a promising transport technique in BISDN. In ATM, data from a traffic source is segmented into fixed-size cells (53 bytes). Due to the small ATM cells, some ATM traffic sources are modeled as a periodic bursty on-off source. In a bursty traffic, it generates cells at peak rate during burst (on) period, and on the other hand no cells are generated during silent (off) period. If we consider the number of active sources during a frame period (D) at burst level, the state of arriving traffic sources is characterized by discrete-time Markov birth-and-death (BD) process. Due to the statistical variations in traffic, there may be a congestion in the ATM network. When a congestion occurs, the cell delay and loss increase, which degrade the quality of service (QOS). One important solution to this problem is call admission control (CAC). The call admission control (CAC) is the mechanism that decides whether or not a new connection can be admitted to the network. It attempts to assure the connection can be accepted without compromising the required QOS requirement of the new connection request as well as established connections. Once admitted, the actual traffic from a user is monitored to determine whether the arriving traffic conforms to the declared traffic parameters. When the traffic exceeds the values of declared parameters, the excess portion of the traffic from the source is policed by the punishment, tagging or discarding. This thesis proposes a burst level admission control approach, which computes a cell loss rate (CLR) in the discrete-time Markov BD process, and the CLR obtained is compared with the cell loss QOS. CLR is computed from the expected path length derived by modifying the concept of a conventional first passage-time. The First Up-Passage Time (FUT) and the First Down-passage Time (FDT) are introduced. From these concepts, the algorithm to compute loss period and CLR is proposed. The performance of our CLR computation algorithm is compared with others. It shows an outstanding features in computational cost in real application.