HiCom: A Hyper-ICN Architecture for Computing Power Network in Edge

The ubiquitous computing power resources of diverse devices and platforms are expected to be interconnected to deal with the rapidly increasing data processing demand at the network edge, whereas they face essential challenges such as computing resource discovery and management. In this article, we propose HiCom, a novel Computing Power Network (CPN) architecture based on Information-Centric Networking (ICN), to enable edge servers and end devices to complete computing tasks on demand in a cooperative, automated, and dynamic manner. Specifically, HiCom features three innovative designs. Firstly, HiCom leverages integrated yet differentiated management of heterogeneous CPN resources in the control plane, by utilizing a multi-dimensional naming space to match different resource features. Secondly, HiCom ensures prompt result returns despite uncertainties in service processing times and transmission path variations, by integrating IP’s destination-determined routing into the service mechanism. Thirdly, HiCom supports computing tasks with data dependencies of sub-services through multi-server coordination, by employing a Directed Acyclic Graph (DAG)-based scheduling mechanism. The developed prototype demonstrates HiCom’s advantages over representative IP-based and ICN-based schemes, including enhanced stability and lower latency for task execution in dynamic edge scenarios.