Section: New Results
Efficient switched services over a DVB-SH / 3G network
Participants : Azza Jedidi, Frédéric Weis [ contact ] .
The architecture studied in the scope of this work is based on a unidirectional DVB-SH broadcast network, coupled with a third generation cellular network. DVB-SH provides users with a variety of services, which could be classified in several categories. It offers real-time applications. Examples are TV-like broadcasting, live broadcasting and notification, which consists in broadcast notifications sent according to the preferences of the user (notifying a football fan of the retransmission of his preferred team matches for instance) and games, like real-time quizzes or multiplayer online role-playing games, etc. It also provides applications to download. For large general audiences, data file purchase services are offered, either on a subscription basis, such as downloading every morning the electronic version of the user's newspaper, or on an impulsive purchase basis, like for films, books and audio CD purchase.
Besides, one of the main characteristics of the Internet world is its bidirectionality, permitting full interactivity to users. In this work, a DVB-SH broadcast network is combined with a third generation cellular network (3G network) to ensure this bidirectionality, as shown in Figure 3 . Actually, this convergence takes benefit from 3G and DVB networks. 3G network characteristics, especially upload link, enable added-value services and applications that are interactive and more personalized. DVB-SH is provided with a very high bandwidth capacity that allows unidirectional IP-TV channels broadcast. We identified two scenarios of services that could be realized:
DVB-SH offers an important broadcast capacity. A residual bandwidth in the DVB-SH path may still be available because of the variable bit rates of served flows. Our first service scenario focuses on this small residual bandwidth and its potential utilization. The idea is to realize an efficient switching of IP popular services, coming from 3G networks, to the residual bandwidth of DVB networks. The goal is to use the architecture in order to provide interactive low cost services over DVB networks.
The second scenario considers classical DVB channels, including TV programs and advertisements spots. We enhanced such a TV service through the definition of personalized advertisements spots, that better fit user interests and localization. Obviously, advertisements are here given as an example of personalized service, but other types of contents could be proposed. In this scenario, the personalized content is sent over 3G, while the DVB content is still broadcasted. The terminal entity receives both contents, but plays only the personalized one. An important issue here will be the synchronization of flows to be read.
Coupling 3G with DVB networks for scheduled and non scheduled services
A residual bandwidth in the DVB-SH path may still be available because of the variable bit rates of served flows as shown in Figure 4 . We focused on this small residual bandwidth and its potential utilization. The idea was to realize an efficient switching of IP data, coming from 3G networks, to the residual bandwidth of DVB Networks. The goal was to use this architecture in order to provide interactive low cost services over DVB networks.
Within this scenario, we proposed to manage two kind of 3G services: (1) scheduled services and (2) non scheduled services .
A scheduled service is a service whose availability is fixed and announced in advance. Users can subscribe in advance to the service. An important consequence, is that we know in advance if such a service is popular or not, thanks the number of subscriptions. 3G service providers ask for broadcasting their popular scheduled services over the DVB residual bandwidth. In a first step, we switched and inserted the 3G popular flows over DVB-SH residual bandwidth using a simple First In First Out (FIFO) policy. Then we refined our flow switching model through the definition of more adapted scheduling algorithms. Using Simulations, we showed that the delivery was a little bit delayed as the residual bandwidth is very small, however the service is efficient as it replaces numerous unicast connexions by a broadcast delivery over a bandwidth initially unused. Moreover, user are satisfied as such a service is offered at very low cost.
In a second step, we proposed to extend DVB residual bandwidth. Actually, 3G provider can pay for reserving some DVB channels. Our objective was to study the impact of this extension on the type and the quality of targeted services. This additive residual bandwidth allows to decrease the service transmission delay, and to implement "non scheduled services" not limited by availability announcement constraints. In this context, our study presented a user subscription prediction algorithm that aims to guarantee switching efficiency while maintaining service continuity. Thanks to bandwidth extension, the delivery delays have been reduced and our service could target a larger scope of services.
Customized content services
In this second scenario, we still considered a DVB-SH / 3G architecture, presented in figure 3 . DVB-SH network delivers several TV channels. Each TV channel proposes many programs, which are broadcasted to all subscribers. The goal of this scenario was to enhance the offered service, by adapting some contents to make them fit user's interests and actual needs. Recent researches have considered the personalization of services and have proposed material solutions to provide regional content through DVB.
Our approach was different. Actually, we took benefit of the coupling between DVB and 3G path to provide personalized DVB TV contents. During the subscription process, users deliver their user profile, which indicates their main interests. During advertisement sequences, they continue receiving the DVB content, but, they also receive the 3G customized advertisement. Running some simulations, we observed that our scenario was not scalable as it involved too many unicast connexions thus prejudicing the performances of 3G "legacy" services. As a consequence, we changed our criteria of service personalization. Rather than using the user main interests based on its profile, we used user location. Actually, users sharing the same location have often similar interests. Users attending to a music concert are interested on music advertisements. Users attending to a football game are interested on football or other sports events, etc. We finally proposed to replace the 3G path by an MBMS network. MBMS environment provides the necessary tools to broadcast a unique content to all users located in the same radio cell. A different content can be delivered in each cell.