Internet today is a best-effort network, with time-varying bandwith characteristics. A system for multicasting video over the Internet has to deal with heterogeneity of the receiver's capability and/or requirements. So adaptive mechanisms are needed. Real-time multimedia traffic places a number of constraints on the network. Conventionally multi-layered transmission of data is preferred to solve the problem of varying bandwidth in multimedia multicast application. In todays scenario majority of the flows are highly bandwidth consuming and are bursty in nature.
Consequently we observe sudden changes in available bandwidth, leading to lot of variations in received video quality. Our aim is to minimize these variations. In our approach, throughtout transmission, we have a notion of available bandwidth using ``bandwidth prediction model". This model is refined periodically using feedback from receiver. We also propose to utilize ``startup latency" i.e. the time before playout starts at client,to overcome the problem of bandwidth variation that may arise later.