IFCNS: Intelligent Forwarding and Computation Node Selection in ICN Mobile Edge Computing

Authors:
Publisher website
Google Scholar
Abstract
Information-centric networking (ICN) is an emerging network architecture for mobile edge computing (MEC) referred to as ICN-MEC. ICN-MEC has garnered significant attention for its capability to offload compute-intensive applications that utilize microservice architecture to edge servers. For such compute-intensive applications, these microservices depend on each other, are known as dependent microservices, and must be sorted in a specific sequence order by solving their dependency relationships before offloading and execution. However, such sequence orders result in waiting time for certain dependent microservices. We introduce Intelligent Forwarding and Computation Node Selection in ICN-MEC (IFCNS) to address this waiting time for offloading dependent microservices. IFCNS addresses microservice dependencies by partitioning them into distinct types and designing a naming scheme extractable by intermediate forwarders. The intermediate forwarder also maintains records of edge server computation loads, predicted using deep neural networks (DNN), and selects suitable edge servers for offloading. Acting as a decision-maker, the intermediate forwarder makes intelligent offloading decisions to reduce waiting time. Extensive simulations using the ns3-based ndnSIM simulator show that IFCNS significantly reduces time consumption by approximately 60%-70%, backhaul traffic load by 57%-68%, bandwidth consumption by 30%-40%, time consumption overhead by 75%, and energy consumption by one-third compared to the DVR-MPTS and OTOOA schemes, respectively. Furthermore, IFCNS enhances computation request satisfaction efficiency by 9% and 22.87%, respectively, while reducing computational resource utilization by 10%-90% compared to the aforementioned schemes across varying computational request rates.
Funding Information
  • National Research Foundation of Korea (NRF) grant
  • Korean government (RS-2022-NR069069)

This publication has 46 references indexed in Scilit: