Voice/Video over IP (VoIP) systems to date have been either highly centralized or dependent on the IP multicast in nature. Global Multimedia Collaboration System is a scalable, integrated and service-oriented VoIP conferencing system, based on the XGSP collaboration framework and NaradaBrokering messaging middleware. This system can provide media and session services to heterogeneous endpoints such as H.323, SIP, Access Grid, RealPlayer as well as cellular phone. In this paper, we address the challenges of scalability, interoperablity and heterogeneity in massive VoIP conferencing system. We believe that our approach opens up new opportunities for leveraging classic VoIP systems by using new technologies in service-oriented computing. Keywords Videoconference, Service-Oriented Architecture, Global-MMCS, XGSP, NaradaBrokering,