A proactive scheme using efficient preemptive route maintenance for Ad Hoc on-demand distance vector (AODV) routing

Abstract

In this thesis we propose a proactive scheme using efficient preemptive route maintenance for AODV (Ad hoc On-demand Distance Vector) routing. Existing on-demand routing algorithms such as DSR (Dynamic Source Routing) and AODV do not consider maintenance of established paths. So after a link failure is detected, the path discovery operation is triggered. This causes delay and jitter to packets delivered. To address this problem, the recovery action is triggered early in preemptive route maintenance by detecting that a link is likely to break soon and it finds an alternative path. But preemptive route maintenance by broadcasting RREQs to find alternate good path can degrade performance of total network in terms of throughput, end-to-end delay as network topology is changed frequently and traffic flows increase. Motivated by this point, we propose a new proactive scheme (GPAODV) in which nodes in the near of destination generate GREPs (gratuitous replies) to build extended sub-paths to source and the generated GREPs are forwarded by "greedy forwarding" strategy using knowledge of position information of neighbors and source. For this, we assume all data packet convey position information of their sources and hello messages position information of their originators. Once extended sub-paths are established in this way, they are maintained by a proactive manner: explicit periodic hellos and errors. In alternative path discovery operation, we introduce the alternative path request (APREQ) message to request delayed reply (DREP) and only good route replies (RREP) in terms of hop count. We believe that by this approach, delay and overhead for establishing alternative paths can be reduced. We implemented our scheme by extending AODV implementation in NS-2. Also, we implemented "emulated" preemptive route maintenance (PAODV) for the comparison with our scheme called GPAODV. The simulation results prove that our idea is more efficient than PAODV in terms of num...