QoS measurement method, measurement setup

As we have seen, the wireless network will work efficiently if the traffic is localised. This can be achieved by providing an out of band connection (wired or wireless) to node groups frequently enough to minimise the number of hops between the communication entries. From the literature we have learnt that the maximal hop count is nearly four hops. In this case, with a tolerable error, the WMN can be modelled as a four hop chain of nodes. With this model we ignored the transversal traffic, but as the traffic is heading toward high bandwidth connections, the model can provide us with information about the QoS services of a WMN with sufficient precision.

We have built a test bed to measure the QoS and the raw data transfer capability of four hop long chain of wireless nodes. We have measured the QoS parameters with different codec types with a compressed real-time protocol and with a traditional real-time protocol. We have tested the effect of the voice activity detection on the QoS of the communication channel. The measurement has been conducted both with and without background traffic. At the access points we have used the Freifunk [40] firmware and the measurements have been executed by the D-ITG [41] software package. To avoid the interference from other 802.11 sources the measurements have been realized in the basement of our building. The topology has been enforced by setting the level of transmit power on each access point to the minimal level at which the neighbours have been able to hear each other. The clock of the server and the client have been synchronized with the help of Windows time services.

Figure 1.

Table 1.

The results of the measurement are shown in table 2. The CRTP has not provided significant effect therefore it is not presented in the table. The number of hops has increased the delay but only by milliseconds depending on the data rate of the connection. As it had been assumed, the higher data rate provides better delay and jitter values. There has been no packet loss for the first two hops. On the third hop it has been about 0,1%. Based on the G.114 [55] recommendation the upper bound of the total delay should be less than 150 ms, so from these results we can conclude that only 802.11g based WMN can provide acceptable QoS for data paths with larger (>3) hop count.

The following table shows the effects of UDP background traffic (5 MBit/s). We have measured this without the support of the QoS of the devices. The long delays show that for a VoIP enabled WMN the QoS management is of critical importance. We should take care about the QoS not only by hop-by-hop basis but it should be an integrated approach considering the full path.

Table 2.