Key points are not available for this paper at this time.
Currently, data processing employs a three-tier computing power architecture encompassing cloud, edge, and end. However, under such an architecture, large-scale, ubiquitous and heterogeneous computing resources distributed in the cloud, edge, and end are isolated from each other, resulting in low resource utilization and poor service performance. Computing Power Networks (CPNs) use the network to connect distributed computing resources to realize the deep integration of computing and networking and provide users with unified computing services through integrated resource orchestration and network control, thus offering a promising solution. Task scheduling is a pivotal technique in CPNs, as it is closely related to the quality of user experience. However, existing studies on task scheduling focus on the selection of computing nodes while ignoring the scheduling of the network, and most of them are unaware of the underlying service intent of applications during the scheduling process. Toward this end, drawing on the idea of Intent-based Networking (IBN), this article proposes a Service Intent-aware Task Scheduling (SIaTS) framework for CPNs. The computing power identification method and the service intent-aware mechanism are designed. An auction-based task scheduling algorithm is developed to achieve the optimal matching of task intent and CPN resources. Numerical results evaluate the performance of the proposed SIaTS.
Tang et al. (Fri,) studied this question.