Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

92 P.M. Ruiz et al.MIGAccess NetworkMMARP_3MMARP_2MMARP_1WWRWRSenderFig. 3. Topology of the testbedthe network depicted in Fig. 3, in which every MMARP node is switched off, sothat there is a dedicated IEEE 802.11b wireless link between the receiver and theWWR. The wired part of the network is running the PIM-SM routing protocolto create the multicast path between the source and the WWR, which acts as anIGMP designated router forwarding multicast datagrams to the receiver whenit joins the source.The multihop tests are exactly the same regarding the wired part of thenetwork. However, we deploy a self-organising ad hoc network extension withnodes running MMARP rather than a single-hop link between the receiver andthe WWR. The receiver joins the multicast source in the fixed network throughthis multihop access network.We use CBR traffic generator to measure the end-to-end bandwidth andpacket delivery ratio. This application generates UDP packets with a payload of900 bytes (i.e. 942 bytes including the IPv4 and UDP headers) which are thenaccounted at the destination. For each of the tests we have performed severalmeasurements at increasing distances (7m, 15m, 24m, 30m, 42m) between thereceiver and the WWR. At each distance, we have repeated the measurementsusing three different data rates of 100 packets/s (753.6 Kb/s), 50 packets/s (376.8Kb/s) and 25 packets/s (188.4 Kb/s) respectively. The results of the differenttrials are described in the next section.As expected, in our indoor scenario the performance depends not only on thedistance but on the node’s position as well. This is mainly due to random noisecaused by traversing walls, obstacles, etc. The results are calculated as the meanvalues over quite a huge number of measurements per experiment. In addition,the measurements are performed in the same positions for each distance. So,random noise is expected to be nearly the same in all the trials, not affectingthe validity of our experiments.4.3 Experimental ResultsTo be sure about the cause of the packet losses in our analysis, we empiricallychecked that there were not packet losses within the wired part of the network.So, all the packet losses perceived at the receiver will occur in the wireless partof the network.