Section: New Results

Multicast Routing

Participants : Ali Boudani, Bernard Cousin, Thierry Feuzeu, Alexandre Guitton, Raymond Marie, Miklós Molnár, Joanna Moulierac.

Scalability and efficiency of routing, QoS management and tree construction have to be improved for multicast traffic. For a global description of our contributions, see [26] and [23] . Our research on multicasting can be organized in several thematic axis:

Tree Construction

In optical networks the constraints for routing algorithms follow from the physical constraints of optical links and switches. One of the fact which restricts well known multicast routing algorithms in WDM networks is that the splitting capability of messages is missing. In order to adapt multicast routing algorithms to WDM networks and improve their efficiency, a new way of doing multicast on wavelength-routed optical networks was proposed. A summary can be found in [10] .

Digital TV, IP and MPLS Multicasting

Our research has been deployed in the domain of digital TV broadcasting. More specifically we have proposed an effective way to transfer TV media in a unified framework. This framework enables the easy integration and management of distributed broadcasting systems which are built with heterogeneous equipment [54] . The goal of this work is to specify an optimised transfer mechanism, this mechanism is integrated in a unified framework managing digital TV content. Our transfer mechanism provides an abstraction level for different network infrastructures and protocols in a seamless way, taking advantage of each of these technologies. For example, multicast and QoS features are supported and managed, as far as they are provided by the underlying network interfaces. Moreover, striped and partial Media Asset transfers are handled independently of the network infrastructures, at the Application layer.

We have proposed several enhancements for efficient management of multicast tree either for plain Internet data network [99] , [24] or for internetworking over MPLS infrastructure [99] , [50] . Using Internet over MPLS provide the efficiency and traffic engineering capability given by label switching and the flexibility given by Internet. In this context we have proposed to detect branching nodes and to use LSP to minimize the working load induced by multicast routing. A simulation platform has been extended to analyze the multicast traffic over MPLS.

Explicit Group Multicast

We have enhanced Explicit Routing Technique, which enables routing of small groups on Internet. Our proposals decrease packet overhead, enable management of larger groups, and segment efficiently the packet. We have proposed an extension to the Xcast protocol [116] . Our proposition is an adaptive protocol which generalizes the Xcast forwarding method and may be parameterized to fit the size of the group. Our protocol is as efficient with small groups as the original Xcast family protocols and it can manage more efficiently larger groups [115] . This solution has been proposed in an IETF draft .

Xcast like explicit multicast routing procedures implicate the overload in the concerned routers. Tree based explicit routing solutions can appease the overload with explicit coding of the diffusion tree. The actually known tree based explicit multicast routing protocols do not handle the fragmentation of forwarded packets. We proposed a new tree based explicit multicast routing protocol which can handle the packet fragmentation and which permits the use of the tree based routing even if the multicast group is large. A first description of the routing problem and the proposed solutions is presented in [10] .

Label Switching

We propose a bridging method based on label switching. This method produces no overhead: labels are located into the Destination Address header field, but gives high flexibility. This flexibility can be used to efficiency manage QoS or multicast traffic. A prototype is developed to measure the performance of our proposition [61] .

Multicast Tree Aggregation

Multicast is not scalable in IP due to the number of forwarding states needed in routers. The aggregation of the used multicast trees is a recent proposition to decrease the number of used trees and so the number of forwarding states in the routers. The previously proposed algorithms for tree aggregation need important computation capacities and time. To simplify the tree aggregation a fast algorithm was proposed in [67] . Generally, the aggregation algorithms do not handle the QoS needs of multicast communication requests. A solution to handle the bandwidth environment of multicast communications was elaborated in [86] .