Dynamic Clustering for Low-Delay Delivery of Video Content Cached in MEC Servers

Abstract

For the purpose of video caching and low-delay video delivery to the end-users, multiaccess edge computing (MEC) servers are commonly grouped into clusters to efficiently exploit the limited storage resources of the MEC servers. In this article, we first introduce a methodology for analysis of the video delivery delay using the queuing theory. Our analysis shows that the video delivery delay is mainly affected by the arrival rate of the video requests. Furthermore, we show a tradeoff between the ratio of videos found in the MEC servers within the same cluster and the transmission delay of video contents. To reduce the video delivery delay, we propose a dynamic MEC server clustering (DyMECC) algorithm that determines the cluster size at each time interval by solving analytically derived equations considering the actual arrival rate of video requests. It also acts as a congestion avoidance mechanism for the communication interfaces among the MEC servers. Via simulations, we show that the DyMECC reduces the video delivery delay about 15% in light load conditions and by more than five times in heavy load conditions compared to state-of-the-art works while also reduces the load of the communication interfaces among the MEC servers by more than 45%.