Efficient Transfer Method for On-demand Video Delivery Based on Streaming Packet Analysis

This paper proposes an efficient multicast-like transfer mechanism for on-demand video streaming. The rapid growth of Internet video services such as YouTube has caused increased traffic and network congestion. Network driven multicast-like transfer is a desirable protocol to suppress unicast-based video traffic efficiently using multicastlike packet forwarding. To analyze the applicability of this mechanism to on-demand video delivery services, this paper first confirms the packet sameness when multiple users virtually simultaneously watch the same video contents. Based on the results from packet-based inspection, we then propose the video packet synchronizer and the video packet buffers and implement them for one of the conventional multicastlike transfer protocols. The experimental results show that the proposed network-driven multicast-like transfer scheme can reduce up to 25 percent of video traffic.