This paper introduces a new Sender-Adaptive & Receiver driven Layered Multicasting (SARLM) scheme for layered or scalable video. Sender first codes the video sequence into multiple data streams by using scalable source codec with FEC (Forward Error Correction) and then sends each stream as a separate multicast group. Receivers dynamically estimate the available bandwidth using packet-pair approach. They join or leave groups independently according to network condition. Meanwhile, receivers send sparse feedbacks containing statistical information about corresponding network status. According to the analysis of feedback messages, sender adjusts sending parameters. Simulation results show that higher throughput and better quality can be achieved using SARLM.