Dual-label aware service replacement for interaction quality improvement in heterogeneous MEC system

摘要

Interaction quality is an important factor for service provision to achieve better user experience. Mobile Edge Computing (MEC) is a promising paradigm to improve interaction quality by supporting near data computing at the edges. However, the limited resources at the edge nodes make it hard to response various services simultaneously, while the service load changes over time. Hence, it is important and challengeable to utilize the limited edge resources to host various service and reduce service response time to improve interaction quality. In this paper, we investigate the service replacement problem to adjust the edge resource utilization dynamically and then reduce the service response time. We first propose a priority placement (2P) algorithm to place the services among the edges by taking account the service priority, which indicates the influence for response time reduction. Then, we propose a dual-label aware service replacement (D-LASR) algorithm to achieve dynamic service placement to fit the service load variation. The replacement strategy works based on the delay sensitivity label and the load gradient label, which represent the features how the service location and service load affect the service response time. We conduct extensive simulations and the experimental results show that the D-LASR algorithm can reduce the average service response time by 40-60%, which indicates that the D-LASR algorithm has better performance in improving interaction quality for service provision in MEC system.