Abstract

Conferencing enables the conversational exchange of media between several parties. Conferencing applications are among important enterprise applications nowadays. However, fine grained scalability and elasticity remain quite elusive for multimedia conferencing applications, although they are key to efficiency in the resource usage.
Cloud computing is an emerging paradigm for provisioning network, storage, and computing resources on demand using a pay-per-use model. Cloud-based conferencing services can inherent several benefits such as resource usage efficiency, scalability and easy introduction of different types of conferences.
This thesis relies on a recently proposed business model for cloud-based conferencing. The model has the following roles: conferencing substrate provider, conferencing infrastructure provider, conferencing platform provider, conferencing service provider, and broker. Conferencing substrates are generally atomic and served as elementary building blocks (e.g. signalling, mixing) of conferencing applications. They can be virtualized and shared among several conferencing applications for resource efficiency purposes. Multiple conferencing substrates provided by different conferencing substrate providers can be combined to build a conferencing service (e.g. a dial-out signalling substrate and an audio mixer substrate can be composed to build a dial-out audio conference service).
This thesis focuses on the conferencing infrastructure provider and conferencing substrate provider roles. It proposes a virtualized cloud infrastructure for multimedia conferencing applications. This infrastructure relies on fine grained conferencing substrates (e.g. dial-out signalling, dial-in signalling, audio mixer, video mixer, floor control, etc.) and offers several advantages in addition to fine grained scalability and elasticity (e.g. assembling substrates on the fly to build new conferencing applications). An architecture is proposed to realize the roles of conferencing infrastructure provider, conferencing substrate provider and their interactions. A resource allocation mechanism for conferencing substrates is also proposed. We have also built a prototype with Xen as the virtualization platform and validated the architecture. Performance has also been evaluated.