1 Analytical Evaluation and Dimensioning of a Cross-Layer Link Sensing Algorithm Husnain Mansoor Ali, Anthony Busson, Saioa Ros Jim´ enez, V´ eronique V` eque Abstract—In this paper we present and analyze an algo- rithm aimed at link management for mobile ad hoc net- works. There are frequent link failures in an ad hoc net- work especially if there is high mobility of nodes. Routing protocols thus have to ensure that the path used for com- munication is not outdated. For this, different approaches have been proposed. Our proposed mechanism improves the techniques which are used nowadays and is based on signal strength measurements. It takes into account tem- poral as well as signal strength stability of a link. These information help in anticipating link breakages and guaran- tees that only those links that have good quality are used for communication. This makes the link management more robust thereby greatly improving the performance. Index Terms—Ad-hoc Networks, Routing Protocols, Link Management, OLSR, MANET, Signal Strength I. Introduction The Optimized Link State Routing Protocol (OLSR [5]) is a table-driven pro-active protocol for ad hoc networks. Being pro-active, each node maintains information about all the other nodes at all times. This information is gath- ered and updated based on received control traffic. This control traffic consists mainly of TC messages which con- tain a more or less exhaustive list of the neighbors of a node. The TC message is broad-casted hence all nodes re- ceive the TC messages sent by a node. This allows each node to know the entire topology and to compute the path for each destination. In this context, the accuracy of the routing tables is highly dependant on each node accurately knowing its neighborhood. Normally, OLSR uses three- way handshake to establish a link and uses link validity times for link breakage. Hysteresis is used by OLSR to improve link management by slowing down connection es- tablishment and speeding up breakage. Hysteresis, thus provides more robust link-sensing which, in our opinion, is necessary for providing better performance when consider- ing high mobility scenarios. The strategy used by OLSR, or more generally defined in the Next Hop Discovery Protocol [4] is based upon two functions, stability rule and instability rule, and uses two link-quality thresholds (HYST THRESHOLD HIGH and HYST THRESHOLD LOW). Hysteresis requires a node to maintain a link-quality value for each link. OLSR RFC specifies hysteresis based on HELLO packets. The stability rule is applied on a registered link every time a hello packet is received at that link. The instability rule is applied to a This work is supported by the French government funded project ANR RNRT R2M (Reseaux Mesh et Mobilite - Mesh network and mobility) University Paris XI IEF - CNRS UMR 8622, Centre Scientifique d’Orsay, 91405 Orsay - France, email:{husnain.ali,anthony.busson, saioa.rosai,veronique.veque}@ief.u-psud.fr registered link every time a hello packet is lost which is de- tected by using interval timers and also by tracking packet sequence numbers to find missing packets. The status of a link is only changed if the link-quality crosses one of the two thresholds. The above strategy while beneficial in low mobility sce- narios, doesnot work well when the speed of nodes are high. It fails since the time taken to establish and break links be- comes significant.This leads to the loss of data packets sent on links which are no longer valid. To overcome this, we proposed in a recent paper a new algorithm to manage links more efficiently in presence of high mobility [1]. We modified the OLSR Hysteresis such that instead of being based just on Hello losses, its now based on combination of Hello losses and received signal power. By using signal power, we also gain the advan- tage of anticipating link breakages. This algorithm is de- signed for mesh networks where part of the network nodes are fixed or have little mobility and are dedicated to for- warding packets (see [6]). The performance of the link sensing algorithm is thus dependant on the dimensioning of the stationary/fixed nodes. In this paper, we present dimensioning rules and evaluate the performance of the algorithm. In the next Section, we present briefly the link sensing algorithm. In Section III, we derive the dimensioning rules of the algorithm. In Section IV, we compare via simu- lations the performance of the native OLSR link sensing algorithm with our proposed algorithm. Test results stem- ming from an implementation of the algorithm is presented in Section V. We conclude in Section VI. II. Link Sensing Algorithm The goal of our algorithm, named Hysteresis on Signal, is to progressively increase the link-quality metric when a node’s signal strength is increasing (node is approaching) and to decrease the metric when signal strength decreases (node is moving away). For this, we use two thresholds ss threshold low and ss threshold high. Only the links for which the power is above ss threshold high are considered by the routing protocol. So, when the signal strength of a hello packet is above ss threshold high, the hello is con- sidered as received and stability rule is applied. When the signal strength is below ss threshold low, the hello packet is supposed to be lost and the link quality is decreased by applying instability rule. When the received power is in between the two thresholds, the nodes may be moving away or approaching. In this case, we compare the signal strength of the successive hellos to anticipate link breakage or link establishment. In an ideal situation, the link-quality