Our protocol view of the wireless architecture

In our development we have focussed on such a wireless NGN and an associated service architecture where adaptive applications offering Web- and real-time multi-media services are running above a QoS-control, service-, location- and security-management layer on top of the transport and network layers with their distributed resource allocation and mobility-management functionalities (see Fig. 2).

Figure 2: Protocol stack and programming model of the NGN architecture.

The latter comprise basic functions supporting the registration and authentication of users, the attachment of their terminals to access points (APs) including the potential handover among these APs, as well as monitoring and resource reservation functions. These functionalities provide the means to adapt the applications dynamically to the changing conditions of the network environment during the movement of a terminal and its user, e.g. the adaptation of the source rates of a real-time application to the variable transport capacity subject to the changing transport error and overload conditions and the permanent adjustment of the emission rates of sources of elastic services by optimized flow-control, media conversion and adaptation procedures (cf. [26]).
Considering the packet transfer at the flow level, a monitoring of connections and their associated flows is used as basic tool to adjust the resource-reservation and mobility-management processes. The corresponding resource-management agents are placed near the mobility agents to guarantee a simplified information exchange and interworking. By these means mesurement-based CAC functionality combined with the adjustment of flow specifications provides the basic resource control and signaling mechanism (cf. [19], [20]). Extending the Intserv service model, we follow in our approach the concept of a hierarchical aggregation of flows and use an intelligent mapping of such flow aggregates arising from a Diffserv model within the wireline core network to those aggregates of the wireless domain. Consequently, seizing of a link spanning the air interface is derived from a simple class-based queueing (CBQ) policy taking into account the terminal mobility and the QoS requirements of different connections offered by mobile-aware services (cf. [15] - see Fig. 5).
Considering the mobility-management functions at layer 3, dynamic address binding is provided by Mobile IP. However, an improved movement-detection scheme based on measurements at the IP layer and messages from the data link layer, e.g. eager or hinted cell switching in an IEEE802.11 WLAN, is required to speed-up the handoff process (cf. [14]). We have not implemented the latter so far but our studies have confirmed the need of such improvements (cf. [22] - see Sec. 3.2).
Regarding the data link layer, apart from a QoS-support by the MAC sublayer for the slot allocation, the most important component is a channel-state adaptation of a WFQ-policy taking into account the on-off error patterns of a wireless link, e.g. wireless fair service or class-based queueing channel-state dependent packet-scheduling (cf. [24]). Unfortunately, in contrast to Hiperlan 2, the current IEEE802.11a/b standards do not support such functionality. However, their extension IEEE802.11e will proceed along that line in the near future and it can be easily integrated into our control framework.
Regarding the network security our model may integrate security mechanisms at different layers of the signaling plane, e.g. authentication techniques during connection set-up, and the transport plane, e.g. wireless equivalent privacy (WEP) in the IEEE802.11 DLC layer, as well as the authentication and encryption provided by IPSec at the flow level in IPv6 or IPv4. Since we consider security issues as important as QoS requirements, they constitute an integral part of our resource-management concept (see Fig. 2). The security layer of our current model is constituted by the WEP approach and the IPSec protocol based on the Free S/WAN implementation (cf. [1]).
To provide a unified concept for the interworking of the resource allocation and traffic-management procedures provided by different layers of the stack, we have developed an abstract programming model (see Fig. 3). In the latter the signaling and control functionality of the session and flow layers is called network interface. Its communication with the controlling resource-management component at the APs of the access infrastructure as well as the data-stream management at the mobile node (MN) is performed by a component called network interface control. Using link metrics such as the signal- or channel-to-interference ratio (SIR or CIR) and link signals of the DLC, measurement functions in the MN are used to indicate movements and to send trigger signals. The latter arise from the changing status of a link (see Figs. 2, 3). In our concept such functions are provided by a monitoring component.

Figure 3: Components of the implemented architecture in the wireless domain.

2002-02-21