KSRCE, Dr.A.Subramani, Professor, MCA Dept - A Proposal on a new ATM switching design

Dr.A.Subramani,Asst.Prof, MCA Department, KSRCE shares one part of his research work with us.

The ATM forum has proposed the Private Network-Network Interface (PNNI) specification to facilitate routing in ATM networks. The pre-path selection or routing algorithms that guarantee a user’s multiple Quality of Service (QOS) requirements remains an open issue (ATM Forum 1996).The PNNI protocol only provides a routing framework but it does not standardize the method of finding an appropriate path for ATM switching design. In this research work a new ATM Switching design with Doubly Finite Queue (DFQ) management for on demand path calculation in congested ATM networks is proposed. Dynamic routing algorithm also called as Multi-Source virtual Dynamic Routing (MSVDR) for ATM networks which is combined with the PNNI protocol is used.
MSVDR uses an adaptive and iterative path search approach and taken advantage of the PNNI hierarchical structure. It consists of the following six major components.
1. Virtual Source / Virtual Designation ATM switching design option for segment-to-segment explicit feedback control through RM cell in PNNI Protocol
2. Source / Designation behavior for on-demand path computation in congested network.
3. Implementation of Route Cache Policy to select the alternative routes to transmit the data.
4. Doubly Finite Queue Management for dynamic data flow control in ATM switches to avoid the data loss and packet discarded.
5. Using MSVDR Rerouting Mechanism to find the shortest path between congested places to destination node.
6. Adaptive Queue Rate Control Algorithm (AQRCA) for regulated the incoming data.

This Thesis guarantees the following properties of DFQ with MSVDR algorithm:

• It uses only short paths: the number of hops of the path used is at most four times as many as the number of hopes of the shortest path algorithm;
• It balances the load: the maximum total size of packets passed on any node is at most three times as much as the optimum.
• It is localized and scales well to large networks: each node only needs information in its local neighborhood to make routing decision; and as a consequence, our algorithm handles dynamic change and mobility efficiently as only a node’s neighborhood is affected.
• It is online: the routing decision of a packet depends only on the previously routed packets, i.e., the current state of the network. It doesn’t need to know the packets in the future.
• Waiting Time: The packet waiting time in queue is reduced three times compare to other queue management.

This research work to reduce the call setup time, minimize the queue length, fast route recovery, maximum link utilization and good throughput compare to the other routing algorithms.