The performance evaluation of a UPC algorithm in ATM networks

Congestion control is a prominent problem requiring a traffic shaping mechanism in Asynchronous Transfer Model (ATM) networks. The method defined for shaping the traffic is called Usage Parameter Control (UPC). UPC is performed at the User-Network Interface (UNI) and at the Network-Network Interface (NNI). The function of UPC is to execute a set of actions, such as discarding or tagging cells, to monitor and control traffic. The Continuous-State leaky bucket algorithm is a very popular traffic-shaping approach for UPC. By sizing the leaky bucket parameters, a switch shapes entering traffic to conform to a "good behavior" contract for the network. In this thesis, simulations are used to study the effect of the leaky bucket method on the cell loss ratio, the throughput, and the delay under low traffic load and high traffic load. The studies are based on negotiated traffic parameters and source traffic characteristics. Also studied are the performances of peak cell rate UPC using this algorithm and the impact of different values of Cell Delay Variation Tolerance (CDVT) on those types of traffic. Experimental results show that there is some degradation of performance of the UPC algorithm under certain bad case scenarios of traffic patterns. Another simulation result is that because a change of CDVT might affect the performance of Generic Cell Rate Algorithm (GCRA), specifying the CDVT at the UNI is a severe problem.