A Conference Control Protocol for Highly Interactive Video-conferencingRuibiao Qiu Fred Kuhns Jerome R. CoxApplied Research LaboratoryDepartment of Computer ScienceWashington UniversitySaint Louis, MO 63130, USAAbstract— Video-conferencing is an efficient tool for distributed collaboration.With the wider availability of broadband wide area networks such as the Internet2,we can expect an increasing demand for video-conferences conducting over broad-band wide area networks. In this paper, we present a conference control protocolfor a highly interactive conference paradigm and its collaboration environment. Insuch collaboration environment, a fixed number of channels (three) for multimediatraffic as well as a common control channel are used. We propose a conference con-trol protocol that uses a three-channel rotation floor control scheme to coordinateaccess to the shared media channels and avoid race conditions. Experimental resultsas well as the implementation in a research video-conferencing system for wide areanetworks show that the proposed protocol can effectively eliminate race conditionswhile maintaining the scalability and reliability .I. INTRODUCTIONVideo-conferencing is an efficient means for distributed collabo-ration especially for people separated by substantial distance. Withthe increasingly pervasiveness of broadband wide area networks suchas the Internet2, we can expect an increasing demand for video-conferencing over these wide area networks in the future.We can identify various paradigms of distributed multimedia col-laboration. These paradigms range from small scale video phoneto highly interactive multi-party video-conferences. They differ intheir degree of interactivity and scalability. Among them, the interac-tive video-conference paradigm requires the highest degree of inter-activity and scalability. A desirable paradigm for interactive video-conferences requires a number of media channels for video and audiostreams from participants. Yet, to achieve good scalability, the num-ber of channels should be limited. Therefore, mechanisms to coordi-nate the access to the shared channels are required. These tasks arecarried out by the conference control protocols.The major functions of a conference control protocol are floor con-trol and session control. Dommel and Garcia-Luna-Aceves [1] identi-fied floor control as the crucial part of interactive multimedia collab-oration, and gave a comprehensive discussion about the issues withfloor control. A floor control framework was described for refer-ence [1]. They also outlined the design issues of floor control pro-tocols in [2]. They also compared various floor control protocolsfor collaborative multimedia environment, and found out that floorcontrol protocols that are based on multicast offer the best efficiencyand scalability [3]. As a generic guideline, no details were presentedfor specific collaboration paradigms though. In the Multi-Flow Con-version Protocol [4] designed for distributed collaboration applica-tions, Yavatkar and Lakshman devised a token-based floor controlscheme. Such scheme can be used in various collaboration paradigmsby different allocation of tokens. However, for interactive video-conference, it can only apply a “strict concurrency control” with theuse of a single token. This implies that only one speaker can trans-mit his media streams at a time, which limits the interactivity. TheConference Control Channel Protocol developed by Handley, Wake-man and Crowcroft [5] uses a shared control channel for managementof conferences ranging from small and tightly-coupled to large andloosely coupled ones. Such shared control channel scheme is adoptedin our proposed conference control protocol.In this paper, we present a conference control protocol intended foran interactive and scalable video-conferencing paradigm. In a collab-oration environment that supports such paradigm, three media chan-nels are used for two interactive speakers. Contention for shared me-dia channels is resolved with a three-channel rotation floor controlscheme. Such a floor control scheme avoids conflicts on shared chan-nels while still maintains the interactivity and scalability. In addition,a dedicated channel is used for out-of-band conference control traf-fic. The proposed conference control protocol is implemented in a re-search video-conferencing system (ALX project [6]) for high qualityvideo-conferencing over the Internet2. The ALX video-conferencingsystem implements the interactive collaboration paradigm, and theproposed protocol has been proved to be able to implement such col-laboration paradigm.The rest of the paper is organized as follows: in Section II, wedescribe a collaboration paradigm for a highly interactive and scal-able video-conference and the collaboration environment that sup-ports such a paradigm. The proposed conference control protocol ispresented in Section III. In Section IV, we show how the protocolbehaves in a real video-conferencing session, and then we concludethe paper in Section V.II. COLLABORATION PARADIGM FOR INTERACTIVE ANDSCALABLE VIDEO-CONFERENCINGA. Collaboration ParadigmsThere are various paradigms for distributed multimedia collab-orations. Some examples are interactive video-conferences, videophone, group meetings, and classroom sessions. All these differentparadigms exhibit different degree of interactivity and scalability. Theparadigm of video phone just involves two participants with no needto scale. The group meeting paradigm involves multimedia streamsfrom participant groups at distributed locations. These streams arepresented simultaneously to participants at each location. There isno need for switching although there are multiple media streams. Ac-cessGrid [7] isa good example of such paradigm. A classroom sessionis a collaboration paradigm that limits the number of media streamsto one at a time. The media stream from a particular participant (theinstructor) is transmitted to all participants during the whole session.Occasionally, the media stream from another participant is allowed totransmit back to the instructor for brief interaction (e.g. making com-ments, or answering questions). In such a paradigm, there is limitednumber of media streams, but no need for switching among mediastreams.In contrast, in an interactive conference paradigm, every partici-pant has an equal opportunity to speak to the other participants, andis able to see the speaker and to hear the ongoing