Description of the testbed nextupprevious
Next:Performance analysis of theUp:Realization of our conceptPrevious:Realization of our concept

Description of the testbed

According to the IEEE802.11b standard the wireless channel provides transfer rates of 1, 2, 5 Mbps and up to 11 Mbps depending on the signal-to-interference (SIR) ratio. In the following, we use the term signal-to-noise (SNR) ratio synonymously if there are no dominating interference effects of other disturbing sources. The access points are connected to our local Intranet by IEEE802.3 compatible Ethernet segments with a transfer rate of 10 Mbps that constitute the distribution system (DS) of the IEEE802.11 architecture (cf. [3]). This Intranet comprises FDDI and fast Ethernet segments providing a transfer rate of 100 Mbps, an access to a Gigabit-Ethernet by a fast Ethernet switch as well as the access to an external 155 Mbps ATM research network by a corresponding ATM switch. Reconfiguring our access network in an appropriate manner, we are able to evaluate the performance and interworking of our WLAN protocols with IPSec, the resource-reservation policy and the flow- and packet-level controls, and to investigate their impact on the data transfer by TCP or UDP and on the adaptive applications.
The protocol context of our testbed comprises the physical and data link layers of an IEEE802.11 compatible WaveLAN supporting channels of 1 up to 11 Mbps over the air-interface. Due to the overhead of the MAC layer the maximal throughput at the network layer is limited, e.g., to 1.810 Mbps for 2 Mbps and Ethernet frames with a maximal size of 1526 byte and according to our experience to approximately 5.5 Mbps on average for the 11 Mbps environment. In accordance with Bing's [7] detailed throughput analysis, in our experiments both the UDP and TCP flows never exceeded these bounds significantly for a persistent period under stationary or roaming conditions (see Figs. 18, 19 - see Sec. 3.2).
As network layer offering mobility-management functions we have used the Mobile IP version 4 (MIP) protocol provided by the Dynamics implementation (version 0.6.2, 11/2000) of the Helsinki University of Technology (HUT) (cf. [16], [25]). The latter provides the standard functionality of MIP to create and manage the interworking of a home agent (HA), foreign agents (FAs) and the mobile nodes (MNs) and some additional optimization options such as a hierarchical structure of foreign agents and the notification of the MIP layer by events of the MAC to improve handoffs. It guarantees a seamless data transfer by TCP or UDP when the MNs are traversing the microcellular structure and have to change their corresponding APs belonging to different subnetworks (see Fig. 9).

Figure 9:Resource reservation under roaming conditions.

Using the possibilities of a network re-configuration described subsequently, we have analyzed in a systematic manner the impact of the terminal mobility on the data flows and the quality of service of the adaptive applications as well as the interworking of the corresponding protocol functions. To study the performance of the data transfer at the network and transport layers and to monitor the status of the wireless infrastructure, we have used a family of different measurement tools including the Distributed Benchmarking System (DBS) developed at the Nara Institute of Science and Technology in Japan (cf. [23]) and the developed enhanced delay-monitoring tool WVPing as well as the status monitor YASM (see Fig. 10 - cf. [22]).

\begin{figure}\quote \small PING 1024 data bytes \\\begin{tabular......-trip (ms) min/avg/max = 17.6/18.1/71.9 (std. deviation = 2.62){ }\end{figure}
Figure 10:Log-Protocol of WVPing.

Next:Performance analysis of theUp:Realization of our conceptPrevious:Realization of our concept