Abstract
In this paper, we investigate the suitability of some protocol stacks (mainly the UDP/IP model) that were firstly proposed for wired networks, to the mobile ad hoc networks (MANETs). These latter, are known to be dynamic as the topology is constantly changing, which affects the validity of the established routes as time tics farther from the start of every routing period. In the UDP/IP model for instance, the IP layer forecasts an IP packet as soon as it receives a data unit coming from the upper layer independently of when this packet will effectively be sent by the lower layer (i.e., the underlying MAC). As such, if a newly created packet using a routing decision relative to its instant of creation, is delayed at the lower layer for some reason, it will be sent using an out of date routing decision and then will wander through established invalid routes leading to poor performances. We first investigate the inherent functioning of such a model (firstly proposed for static networks) and exhibit its drawbacks in a MANETs context. Then, we propose practical enhancements to accommodate a proper behavior suitable to dynamic networks and that firstly provides a priority treatment to control traffic and secondly makes the routing decisions (fills up the IP next hop field) only when the MAC is ready to handle the packet. We conducted an extensive set of simulations to compare both implementations using different load and mobility scenarios and performance metrics but the same routing protocol (OLSR). These simulations show that our proposal solves the malfunctioning of current implementations, yet it allows a better behavior of the OLSR protocol and thus more accurate network performances. (C) 2011 Published by Elsevier Ltd.