1-2hit |
Piya TANTHAWICHIAN Akihiro FUJII Yoshiaki NEMOTO
Major problems of traffic control in ATM networks include how to decide whether a network accepts a new call or not in real time and how to select the best set of Dual Leaky Bucket (DLB) parameter values. To solve these problems, it is necessary to determine the amount of network bandwidth required by the call. In this paper, we present an analysis based on bounding technique to derive an upper bound on bandwidth requirement when the call is characterized by a set of DLB parameters. Consequently, a new definition of the upper bound on bandwidth requirement and simple formulae used for computing the upper bound have been obtained. To clarify the advantages of the derived upper bound, we demonstrate its two applications, one to select the best set of DLB parameter values from candidates for minimizing the amount of bandwidth to be allocated to the call and the other to establish a Connection Admission Control (CAC) scheme. The upper bound-based CAC scheme is fast enough to process in real time due to its simplicity and provides a significant improvement of network utilization compared to the peak rate-based CAC scheme.
Piya TANTHAWICHIAN Akihiro FUJII Yoshiaki NEMOTO
For traffic control in high speed ATM Networks Usage Parameter Control (UPC) plays an important role. The existing UPC schemes have some limitations. It is difficult to implement policy which involves monitoring vioations while guaranteeing QoS for the compliant connections-particularly with respect to bursty traffic sources. This is due to the difficulty in measuring the Sustained Cell Rate (SCR) and Maximum Burst Size (MBS) parameters simultaneously. To ensure prompt action against policy-violations, speedy detection is an important requirement. But the existing UPC schemes do not have a satisfactory response time. In this paper, we propose a new scheme called Markovian State-Dependent UPC schemes (MSDU) to police SCR and MBS parameter violation simultaneously with a satisfactory response time. The MSDU scheme is performed by using two virtual queues: 1) a Markovian State Dependent Service queue and 2) a Fixed Service queue. The discrete time analysis of the MSDU is carried out for a bursty source which is a Markov-Modulated Bernoulli Process (MMBP). The improved effectiveness of the proposed MSDU is clarified by a numerical comparison with UPC based on standard Leaky Bucket scheme.