Design of a new architecture for audioconferences based on SIP Multicast and Genetic Algorithms

Group services like the audioconference require a minimum level of quality of service for the adequate transmission of multimedia packets, and the improvement of signaling processes for the establishment, maintenance and release of multicast group sessions. Traditionally, the SIP (Session Initiation Protocol) servers have dealt with unicast signaling and complementary support IP (Internet Protocol) multicast. The distribution patterns of the session are described on the SDP (Session Description Protocol) payload. As a consequence, the same existing drawbacks of IP multicast affect the performance of SIP-based requested audioconferences. This work proposes a new multicast architecture based on SIP extensions and genetic algorithms in a multicast manager. Such architecture supports overlay multicast. The group management functions are executed inside the multicast manager. Two testbeds were mounted to measure the joining and leaving time for both IP multicast and SIP multicast with several levels of extra load traffic from a potential group member to the source of multimedia traffic without a genetic algorithm as a first approach. The results show that the difference between the minimum and maximum standard deviation value for both joining and leaving time on the architecture proposed with SIP multicast is lower than with IP multicast for the studied scenarios. Finally, we propose a genetic algorithm and describe a case study.