Collaborative edge-network content replication

Abstract

Today's mobile video users have unsatisfactory quality of experience mainly due to the large network distance to the centralized infrastructure. To improve users' quality of experience, content providers are pushing content distribution capacity to the edge-networks. However, existing content replication approaches cannot provide sufficient quality of experience for mobile video delivery. Because they fail to consider the knowledge of user-behavior such as user preference and mobility, which can capture the dynamically changing content popularity. To address the problem, we propose a user-behavior driven collaborative edge-network content replication solution in which user preference and mobility are jointly considered. More specifically, using user-bahavior driven measurement studies of videos and trajectories, we first reveal that both users' intrinsic preferences and mobility patterns play a significant role in edge-network content delivery. Second, based on the measurement insights, it is proposed that a joint user preference- and mobility-based collaborative edge-network content replication solution, namely APRank. It is comprised of preference-based demand prediction to predict the requests of video content, mobility-based collaboration to predict the movement of users across edge access points (APs), and workload-based collaboration to enables collaborative replication across adjacent APs. APRank is able to predict the fine-grained content popularity distribution of each AP, handle the trajectory data sparseness problem, and make dynamic and collaborative content replication for edge APs. Finally, through extensive trace-driven experiments, we demonstrate the effectiveness of our design: APRank achieves 20% less content access latency and 32% less workload against traditional approaches.