Multiparty communication is an emerging phenomenon on the Internet. The envisioned applications include Internet teleconferencing, telemedicine and telesurgery etc. A typical session involves multiple interactions among participants, and requires data transferring between arbitrary pairs of participants. These applications usually have stringent delay requirements, e.g., data delivery must meet individual deadlines, and the shared data must be received by participants within bounded delay variation. If the group is large, group members are scattered far from each other in the Internet, and each member needs a variety of data items, multicasting to the entire group is too costly; On the other hand, if each request is replied independently regardless of other requests, optimizing one request may delay the delivery of the next. Existing approaches that do not consider the interaction of data streams cannot address the delay requirements effectively. This project aims to establish the theoretical basis for performance analysis and seek algorithmic solutions for QoS guarantees on the delay aspect. We will formulate practical QoS problems into new online problems and try to reach some definitive conclusions on QoS issues in multiparty communication. The proposed research will advance the fields of QoS research in networks and online algorithm design/analysis in theory.