IEEE Communications Magazine • June 2004 80 0163-6804/04/$20.00 © 2004 IEEE QOS IN IP AND WIRELESS NETWORKS INTRODUCTION There is a strong consensus today that IP will be the foundation of next-generation networking. Since IP already dominates in the wired world, the research community is targeted toward the deployment of all-IP mobile networks, which not only rely on IP for packet transferring, but also exploit IP-based protocols to perform fundamen- tal operations, like mobility, quality of service (QoS), and media control signaling, among oth- ers. However, direct exploitation of IP mecha- nisms and protocols is not straightforward in a mobile environment, since the latter exposes cer- tain characteristics that impose additional requirements. QoS is still a topic that attains a lot of atten- tion in both the wired and wireless worlds. Regarding the former, the maturity of the inte- grated and differentiated services frameworks has led to a solidification of the research effort. The main Internet Engineering Task Force (IETF) activity in the field is the definition of an end-to-end QoS signaling protocol within the context of the Next Steps in Signaling charter. On the other hand, in the wireless world there is still substantial research activity, primarily stem- ming from its intrinsic characteristics (i.e., the involvement of the radio interface and the impli- cations imposed by mobility). It is evident, however, that the future net- working environment will be strongly character- ized mainly by the heterogeneity of networks, especially regarding the network access part, although having IP as the common denominator. We can envisage (Fig. 1) that the end-to-end path will first traverse one access network that may be a high-speed wired segment (e.g., DSL), a wireless LAN (e.g., 802.11), a wireless WAN (e.g., UMTS), or even a satellite one. These first-hop segments will probably be supported by an IP-based core network, which will at least supply end devices with IP connectivity, and also include a gateway to the Internet backbone. In light of the above, even if every part of the envisaged end-to-end path has its own QoS mechanisms, end-to-end QoS provision is not accomplished easily. The shortage of a standard- ized end-to-end IP-based protocol for establish- ing QoS, the heterogeneity in QoS models 1 that may apply in different domains of the end-to- end path, and the duality of QoS to be achieved (i.e., in both layer 2 and layer 3) contribute to this distressing situation. Standardization seems to be the only approach to ensure end-to-end QoS provision. Taking for granted that different networks or domains are free to follow any QoS model, the focus is turned to the enabling factor for estab- lishing an end-to-end QoS path that consists of different provisioning mechanisms in each Sotiris I. Maniatis, Eugenia G. Nikolouzou, and Iakovos S. Venieris National Technical University of Athens ABSTRACT The emerging next-generation networking environment presents an IP-based core intercon- necting many wireless radio access networks, providing ubiquitous access to end users through a vast variety of wireless devices. Although the IP protocol will be the common denominator, the new environment brings together many dif- ferent interconnecting domains, each following different QoS models, complicating the overall end-to-end QoS process. This article discusses the need to standardize an end-to-end QoS pro- tocol. It does not, however, focus on the signal- ing mechanism, since there is currently a relevant ongoing activity in IETF. Instead, it concentrates on the formulation of the QoS information describing the QoS requirements of the session to be established. It presents the Generic Service Specification Framework that not only enables the QoS requirements of a specific session to be captured (like a generic QoS template), but also the QoS classes of each IP domain can be described according to it. Through the systemat- ic specification of a domain’s QoS classes, an intelligent automatic mapping algorithm can be applied during an end-to-end QoS request, in order to select the most appropriate service class in each domain, as well as to extract the required traffic-related parameters to perform traffic con- trol operations, such as admission control, polic- ing, and scheduling. End-to-End QoS Specification Issues in the Converged All-IP Wired and Wireless Environment 1 The term QoS model is used here to denote a QoS framework or architec- ture, like DiffServ, IntServ, or the UMTS QoS reference model. It refers to both signaling and the service-level semantics.