Wireless Ad Hoc and Sensor Networks

120 Wireless Ad Hoc and Sensor NetworksEquation 3.37 be given by Equation 3.43 with weight tuning provided by Equation3.69. Then the error in buffer occupancy, ek ( ), or packet losses, c(k),approaches asymptotically to zero and the weight estimates are bounded providedEquation 3.70 and Equation 3.71 hold.3.5 ImplementationNetwork simulator 2 (NS-2) is a discrete event simulator targeted at networkingresearch. We have selected NS-2 to implement our scheme. Forconvenience, we denoted our scheme as TQ (transmission-based queue size).3.5.1 NS-2 ImplementationIn NS-2, agents represent endpoints where network-layer packets are constructedor consumed, and are used in the implementation of protocols atvarious layers. We implement our TQ scheme at transport level. The Tqagent represents a TQ sender. It sends data to a Tq sink agent, processes itsacknowledgments, and performs congestion control. It has a separate objectassociated with it which represents an application's demand. TQ assumesthat a packet has been lost (due to congestion) when it observes 3 duplicateACKs and it retransmits the lost packet immediately. The feedback, whichis generated at the egress node, provides the RTT. The following is a UMLstate diagram of the implementation shown in Figure 3.21.The pseudo code for this algorithm is presented below and the implementationis explained after the pseudo code.//pseudo code for TQ algorithm//notation:// Ik:packets arrive rate at the ingress node// Vk:packets departure rate at the ingress node// Uk:feedback variable from the destination// Qlevel:bottleneck queue level// Ir:packets arrive rate at the egress node// Sr:packets departure rate at the egress node// Xk:buffer occupancy at the egress node buffer// Xd:desired buffer occupancy at the egress nodebuffer// kv:diagonal gain matrix// T:measurement interval100 begin