Recently it has been shown that network coding can highly improve the network throughput in the communication systems and specifically increase the downloading performance in P2P networks. This paper presents a new live streaming system for GIA-based P2P networks. In the proposed system the traditional GIA protocol is improved for an efficient use in multimedia applications. We introduce a new mesh-pull layered video streaming framework for GIA-based systems and employ random network coding to improve the network performance. Random network coding algorithm enables each peer to send a single encoded video chunk instead of multiple video blocks toward the destination node and save its downloading and uploading bandwidths to a great extent. Comparison of our new GIA-based live streaming system with the traditional GIA demonstrates that this system can achieve a significant improvement in total end-to-end delay, network throughput and redundancy of bandwidth usage.