Please use this identifier to cite or link to this item:
|Title:||MODELLING A THRESHOLDS BASED DYNAMIC SCHEDULING POLICY FOR ATM STATISTICAL MULTIPLEXER|
|Authors:||Nigam, Manoj Kumar|
|Keywords:||ELECTRONICS AND COMPUTER ENGINEERING;ELECTRONICS AND COMPUTER ENGINEERING;ELECTRONICS AND COMPUTER ENGINEERING;ELECTRONICS AND COMPUTER ENGINEERING|
|Abstract:||The ATM traffic necessarily demands, the use of statistical Multiplexing at the access or transport nodes to facilitate the flexibility in allocation of band-width. This has resulted in considerable research interest in fixing up the scheduling policy and analyzing the performance of a statistical multiplexer. A new scheduling policy for ATM multiplexer with two different queues for real time and non realtime traffic respectively is suggested here. Policy sets a higher priority for realtime traffic and allots a major portion 4S1' of total `S' servers while rest of 'S2 = S-S1' (S1<S2) servers are catered to serve non real queue (queue2). In case of underload, servers can go for natural shift from one queue to another as multiplexer takes packets on priority basis also. But it is only until any queue crosses the threshold. if queue2 crosses it with point `(3' another block of 'S3' servers are shifted from queuel to it, while if queuel crosses a point "a", all `S' servers are shifted to it.The arrival process for each type of traffic is assumed to be MMPP (Maorkovian Modulated Poisson Process) which is correlated and non renewal and closely models the actual random traffic from different sources. The Multiplexer is modeled as synchronous servers system with separate finite buffers for each type of traffic. Delay and drop calculations are analysed using this MMPP arrival process. This is done by setting four parameters namely : Mean message blocking probability of traffic 2 (2); Mean blocking probability of trafficl(2) packets ; Mean c ay of typel(2) packets and Throughput of trafficl(2) packets. This new policy not only gives an immediate attention to real time traffic but also saves non real time to be thrown in negligence in former's presence. This indeed improves the performance of the multiplexer over earlier proposed scheduling policies.|
|Research Supervisor/ Guide:||Joshi, R. C.|
|Appears in Collections:||MASTERS' THESES (E & C)|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.