Collaborative Video Caching In LTE Core Networks

As a representative design of Next Generation Mobile Networks (4G),3GPP LTE has been actively investigated and improved, as well as been widely deployed all around the world in recent years. Along with higher bandwidth and faster data rate provided by4G, online video service are experiencing explosive growth. It is predicted by Cisco White Paper2012, that video-related traffic over the mobile networks will increase16-fold in the next five years, and will take66%of the overall mobile traffic.Different from the former generation of mobile networks, the core network of3GPP LTE employs full IP-packet switching for both voice service and data service, connecting end users with the data networks via the internal protocol of GTP. However, in such "relay" transmission, all requests need to pass the whole core network structure in order to get served. This will bring in unbearable delay and extra network expense, especially for video service with high data-rate requirement. Apparently, video caching, which brings video clips close to the end users, can dramatically improve the Quality of Experience (QoE) as well as save operating expense. We notice that in the hierarchical structure of the LTE core network, serving gateways have strong computing power, providing possibility for in-network video caching. Based on these observations, this work targets the video caching problem in LTE core networks, proposing a collaborative video caching solution, which further includes the following parts:1. Rearch on video placement for in-network video caching. The placement problem is built together with the characteristics of online video requests, aiming to maximize the aggregate hit ratio so as to reduce the amount of traffic incurred by collaborative requests. The formulated problem is proved to be NP-hard and a fast greedy algorithm is developed for it. Using the distributions of online video clips, the proposed algorithm reaches1%～3%of optimal in all cases simulated.2. Research on the collaborative video transmission in LTE core network-s with video caching. Different manners of video transmission in collaborative caching systems are extensively investigated. An approximate algorithm with provable approximate ratio for in-network request routing. In most existing work-s, collaborative requests are considered as static, file-based transmission. This paper studies and compares different approaches of collaborative request schedul-ing, including file-based non-traffic-aware mode, file-based traffic-aware mode, flow-based non-traffic-aware mode and flow-based traffic-aware model. Further investigation over collaborative requests includes server selection and in-network routing. The joint problem is modeled in an optimization framework, and a low complexity algorithm is proposed for it with provable approximation ratio.3. Research on collaborative request routing with joint video adaptation. In practice, video servers use video adaptation techniques, such as scalable video coding and adaptive rate video, to reduce the data rate of each request, so that keep feasible in-network traffic, especially in scenarios with heavy load. This paper aims to maximize the overall user utility while keeping the maximum link congestion under a given upper bound. However, such an optimization problem is proved to be NP-hard. Then a fast algorithm is proposed for it, which outperforms existing solutions and supports much more traffic load.4. A realistic routing protocol is also investigated. According to output of the routing algorithm proposed, a routing protocol is developed in this paper with a hop-by-hop forwarding design. The coordinator computes a flow-splitting database using the proportional flow results from the routing algorithm, and dis-tributes a flow-splitting table to each router. The routers then split each incoming flows according to corresponding flow-splitting table. This protocol is compatible to most existing routers and also supports in further software defined networks, such as OpenFlow.5. Implementation of a prototype caching system of LTE core network. Ex-isting architectures and protocols in3GPP LTE are extensively studied, and a compatible video caching subsystem is then developed, which is further imple-mented within the open core network prototype of OpenEPC.