Contents I. Introduction
Like the steam power in the steam age and the electric power in the electrical age, computing power (CP) has become a new engine of the digital economy age. Whether you are going to a gym, watching a video, or driving a car, every human activity can be empowered by the CP. The rise of intelligent services and applications (e.g., smart city, remote education and AR/VR) has fueled an unprecedented demand for the CP [1]. Computing power network (CPN) has recently emerged to efficiently leverage computing power resources geographically distributed in cloud computing centers, edge computing servers and smart devices. Electric power can be physically scheduled and moved in the power grid. However, the CP only exists in the computing infrastructures and cannot be physically scheduled to its remote users. Instead of scheduling the CP, a CPN should move data from its users to CP nodes (e.g., datacenters and supercomputer centers) where the data can be stored and processed on-the-fly. A fundamental challenge faced by the CPN is how to meet the growing demands for bulk data transfers between its users and CP nodes as well as between CP nodes [2].
