QoS-aware Accounting in Mobile Computing Scenarios Paolo Bellavista, Antonio Corradi, Silvia Vecchi 'LSDUWLPHQWRGL(OHWWURQLFD,QIRUPDWLFD6LVWHPLVWLFD8QLYHUVLW\RI%RORJQD 9LDOH5LVRUJLPHQWR±%RORJQD±,7$/< 3K)D[ ^SEHOODYLVWDDFRUUDGLVYHFFKL`#GHLVXQLERLW Abstract The enlarging market of portable devices and wire- less networks stimulates the provisioning of mobility- enabled Internet services with differentiated levels of Quality of Service (QoS). The finally supplied QoS level can greatly differ from the negotiated one because of the very variable availability of network resources. This significantly impacts on the pricing policies and, conse- quently, on the definition and realization of fair ac- counting strategies. The paper claims the need of mid- dleware solutions able to evolve dynamically depending on user/terminal mobility to monitor, control and regis- ter the finally supplied QoS level directly within the network localities where users/terminals move to. The paper presents the design and implementation of the Active middleware for Quality-aware Accounting of Mobile services (AQuAM) that is capable of accounting final users for the QoS level finally supplied during service provisioning, in contexts of user/terminal mobil- ity.  ,QWURGXFWLRQ Recent advances in mobile telecommunications and portable device miniaturization propose new service Re- cent advances in wireless connectivity and portable de- vices propose new service scenarios where users, access terminals and even service components can geographi- cally move during provisioning. This forces to face sev- eral technical challenges at different levels of abstrac- tion, from the integration of heterogeneous network technologies to user/terminal location tracking, from de- vice-dependent service adaptation to location-aware service provisioning. In particular, service provisioning to mobile devices with wireless connectivity has to con- sider the wide heterogeneity of their hardware/software characteristics and the highly dynamic variations of re- source availability in the different localities that host the roaming terminals [1]. For instance, while approaching a city by train, a tourist with a laptop connected to the large bandwidth network offered to first class seats is likely to receive a high-quality multimedia stream about the historical buildings of the city. Already available Web servers can provide multimedia tourist the information for that service; however, these data should be dynamically fil- tered and adapted to fit both the client location and the bandwidth/visualization capabilities of the access de- vice. On the contrary, while visiting the city downtown, the same tourist can be interested in receiving only fixed images and textual information (maps, small-size pic- tures, historical notes, …) on her wireless personal digital assistant, also to reduce the time/costs of connec- tivity. Or, she would like to receive only textual infor- mation to her WAP-enabled phone, which cannot visu- alize either multimedia streams or fixed images. In any case, mobile scenarios stress the necessity of providing very differentiated Quality of Service (QoS) levels, de- termined dynamically depending on terminal character- istics, user preferences, and local resource availability. The above scenario motivates the re-thinking of service management solutions in general, and, in par- ticular, of accounting mechanisms and tools. Traditional management is tailored to fixed infrastructures of net- work elements and to QoS-insensitive packet communi- cation; it is based on effective solutions and architec- tures to deal with geographic distribution and heteroge- neity of resources and service components. IETF and OSI have proposed management models based on Cli- ent/Server (C/S) interaction and on several variations of it [2, 3]. It is possible to organize hierarchies of C/S components to achieve decentralization and scalability, but the interaction of management entities is usually statically determined for clients and servers in ex- changing remote information. These solutions do not fit well global systems in rapid evolution where the unpre-