This paper proposes an application-level multicast framework based on peer-to-peer communications to improve performance of large scale VOD services. Since the deployment of multicast-enabled network is still not popular nowadays, developing an application-level multicast infrastructure to support multicast delivery would be a good alternative. We address the problems associated with heterogeneous clients and formulate the model of multicast streaming in a VOD service. The proposed multicast framework takes these results into account to reduce server load and network bandwidth. The simulation results demonstrate that our framework can satisfy at lest 40,000 viewers concurrently by utilizing only 70 MB of buffer on each peer node.