Introduction

Currently, the integration of wireless networks into high-speed next generation networks (NGNs) based on the IP technology and its corresponding resource reservation and QoS-management mechanisms is one of the most challenging tasks of network design and engineering. The fast convergence of these wireless and wired networks coincides with the planned deployment of new integrated multi-media and interactive Web services that are independent of the used communication media and the realized mobility patterns of the terminals and users.
Regarding the network design the application services are using new bearer service models like Intserv and Diffserv or their integrated variants with new associated traffic management schemes, e.g. QoS-enhanced OSPF routing, traffic classification, the implementation of per-hop-behavior (PHB) like expedited and assured forwarding and the separation of packet flows by WFQ-scheduling, leaky-bucket type shaping and marking policies as well as advanced buffer management schemes such as RED or RIO (cf. [19], [20]).
Regarding the in-house infrastructure wireless local area networks (WLANs) based on the new standard IEEE802.11 seem to be one of the most promising integration paths to provide a data transfer of 11 Mbps up to 54 Mbps that is required for new Web and multi-media applications (cf. [2], [3]). However, such networks currently guarantee only a best-effort transport of the original data flows based on the TCP or UDP protocols and an IP network layer with mobility support known as Mobile IP (cf. [25] - see also Fig. 2).
In this paper we focus on such a WLAN in infrastructure mode and sketch the concept of a service architecture where adaptive applications are running above a distributed resource allocation, adaptation and QoS-management layer on top of the transport and network layers with their corresponding resource and mobility-management functionalities (see Figs. 1, 2). Our objective is to identify the potential and drawbacks of the current protocol environment and those QoS-support mechanisms applied in a wired IP network. We study ways how the latter can be adapted to the wireless settings to support new adaptive multi-media and interactive Web applications subject to the constraints of terminal and user mobility. For the latter purpose we develop a new programming model and provide a guideline for the implementation of new mobile-aware applications and new resource allocation schemes. Along this line of reasoning, we sketch the complete path from the concept to the realization by a QoS enhanced LINUX implementation at our advanced testbed of an IEEE802.11b compatible WLAN with Mobile IPv4 as network and conventional TCP as transport layer.
Roaming experiments and traffic measurements are used to show the feasibility of the approach and the interworking of flow control, resource reservation, mobility and security management at the different layers of the protocol stack. For this purpose, we analyze the transport performance of TCP and UDP as well as the effects of the changing quality conditions along a transmission path and of roaming both on the transport efficiency and on the resource reservation processes. We also identify the most effective and sensitive control information for the adaptation processes. To achieve these goals, we correlate quality indicators of the IEEE802.11 layers like the signal-to-interference ratio with efficiency indicators of the network and transport layers, e.g. the round-trip delay and throughput of frames, and the dynamics of TCP flow control. Distinguishing the dominant traffic patterns of real-time and elastic data flows, we evaluate the expected impact on the different types of applications.

2002-02-21