Section: Scientific Foundations
System Support for Continuous Ambient Service Delivery
Mobile networks are becoming increasingly heterogeneous. Global coverage is now well provided by 2G and 2.5G cellular systems, and 3G networks (UTMS) are being deployed in some densely populated areas. Nonetheless, very high data rates (WiMax, WiFi ...) will not be available everywhere in the near future, so the delivery of large amounts of information to people on the move will remain limited and expensive. In this context, the main challenge is to provide services as seamlessly as possible.
Pico-cell Architecture
The past few years have witnessed the rise of the cellular networks. These communication systems were designed with a philosophy of any-time any-where service. Users wish to receive and place calls at any location and without delay, to move while talking without interrupting their conversations. This requires ubiquitous coverage, which in turn requires significant infrastructure. A modern cellular system is installed with hundreds of base stations, at a cost of hundreds of millions of euros, in order that a communication link is always available. Such any-time any-where service provision becomes increasingly expensive and suffers from low bandwidth. Covering wide areas with high radio bandwidth requires complex equalization, due to signal attenuation, multi-path fade, and shadowing effects. Sophisticated radio engineering will lead to improved bandwidth, coverage, and mobile access, but this will be expensive, in terms of both capabilities and cost.
In this context, the ACES project has studied an alternate design for wireless networks where intermittent but very high speed is provided to the network through Pico-cells . The latter consists of a set of access points (APs), i.e. antennas around of which are defined radio cells with limited range (about 100 meters). Those antennas are discontinuously spread on the network area, thus providing a many-time many-where service. Actually, the idea of coverage discontinuity brings two major advantages. First, as it implies the use of a fewer number of access points, the architecture deployment will be cheaper. Second, radio cells disjunction hypothesis simplifies the radio frequency band management and avoids interference problems.
Even if this model simplifies network deployment, the connectivity intermittence induces important challenges in order to avoid service disruption. Thus, terminals have to take advantage of the high bandwidth when it is available. For delay tolerant data, a terminal stores data as it passes under a cell. Hence, it may consume the buffered data even when it passes through regions of poor network coverage. Many projects have studied very specific cases where the cells deployment is uniform and data is sent from servers to terminals (down-link). One example of this type of system is studied by WINLAB (Wireless Information Network Laboratory). In this project, cells are equally spaced and the data delivery algorithm is tested in a network with one dimensional system i.e. high ways.
Our approach is based on a more general case in which cells are distributed according to the envisaged traffic and data can be exchanged in both directions down-link (from servers to terminals) and up-link (from terminals to servers). The main challenge is to provide system mechanisms and efficient services addressing the specific constraints of this architecture: discontinuous coverage, user unconstrained mobility, high user density. To cope with the discontinuous coverage of the network, we store data (with caching mechanisms) close to mobile people, just before data delivery. Thus, the placement policy of data within the architecture is conditioned by knowledge of people on the move. The goal here is to define a representation of person mobility in the network architecture, and to use this model for placing data using limited and customised flooding mechanisms.
Through this architecture model, we underline the analogy between heterogeneous mobile networks and multiprocessor architectures (for example the mobile device can be considered as a processor). This approach allows us to map and extend existing caching mechanisms, taking into account the specific constraints of a discontinuous mobile network.
System Suppport in future 4G networks
Interactive IPTV, evolving internet behavior, and more generally new data services (location information services for example) will strongly influence mobile usage. This requires to support high users density at low cost. During the last four years, our first goal was to increase network capacity by using discontinuous coverage combined with system mechanisms (data caching and data distribution). Memory in the network and terminals facilitates service delivery.
It is now possible to go further. The 4G infrastructure operator will mix several technologies: large umbrella cells (3G, Wimax, DVB-based infrastructure), and numerous pico cells. In 4G context, the infrastructure will be much more distributed, and mobile terminals will have to collaborate with several entities in the network to perform service delivery. In other words, mobile terminals will become an active part of a complex information system distributed between several components in the architecture.
We study system mechanisms to improve the terminal's integration in the network: ability to attach simultaneously several networks, capacity to opportunistically push data according to network conditions ...Terminal efficiency will depend on the number of technologies they can work with.
We have started to work on this problem by studying new possibilities offered by a large scale broadcast network coupled with a cellular network. We consider at first the future DVB-SH standard (satellite services to handheld devices) which is an hybrid (satellite/terrestrial) architecture. It is defined as a system for IP based media content and data delivery to handheld terminals, via satellite. Satellite transmission guarantees wide area coverage. Moreover, it is coupled with terrestrial gap fillers assuring service continuity in areas where the satellite signal cannot be received (built-up areas for example). In the context of our future works, a DVB-SH broadcast network is combined with a third generation network. Actually, this convergence will take benefit of 3G network characteristics, especially upload link, to enable added-value services and applications, which will be interactive and more personalized. For example, one could decide to deliver some classical 3G contents over DVB-SH path. This scenario occurs especially when contents or services suddenly become very popular, thus their transmission may take benefit of the large broadcasting capacities offered by a DVB network. The decision to switch a flow to the DVB-SH path could be based on the flow size and nature and on the number of subscribers. The design of dynamic flow insertion over the DVB-SH network involves multiple mechanisms and raises several open issues.
