In-network cache simulations based on a YouTube traffic analysis at the edge network

Abstract

Recently, with the advent of YouTube and similar video streaming websites, online video playback has gained popularity. Further, video traffic, represents the largest portion of Internet traffic. However, online video traffic contains redundant traffic due to identical video accesses. Network virtualization has been studied and developed, which makes it possible to deploy different protocols and new functionalities over the same physical network. in-network processing, the execution of calculation processes on routers, is one of the new features enabled by network virtualization. Further, since YouTube is the largest video publishing service in the world, we analyze YouTube video playbacks at the edge network and investigate redundant traffic and its locality. Based on these recent developments and technology, we propose to reduce redundant video traffic using in-network caching by positing video caches on routers. This cache could be moved to other routers according to users' access. In this paper, we analyze redundant YouTube video accesses and perform in-network cache simulations. According to these simulations, in-network caching can be optimized to reduce not only incoming traffic to the edge network by 42.2%, but also download traffic inside the network by up to 18.9%, with the only cost being an increase of 6.6% additional upload traffic inside the network. The result also demonstrates the presence of the locality of video accesses at the edge network.