Automatic Scheduling Technology of Computing Power Network Driven by Knowledge Graph

Yanheng Bi,Yingchi Long,Yanzheng Jin,Shengwen Zheng,Huaiyuan Liu,Hongzhi Wang

DOI: https://doi.org/10.1109/icss55994.2022.00032

2022-01-01

Abstract:In recent years, the demand for computing resources of AI industry is urgent because of the data explosion, which promoted the construction of computing power networks in the new era for operators. From the cloud network era to today's computing power network, stricter requirements are proposed to ensure the efficiency and security of computing services. Despite computing power scheduling technologies such as on-demand edge computing and efficient compute first network, knowledge graph techniques for graphs are less explored. As a new technology that can express the relationship between nodes in the graph extremely easily, knowledge graph has a natural advantage in expressing feature information of computing nodes in computing power network. Therefore, a novel knowledge graph representation for the architecture of computing power networks is proposed. And the knowledge graph of the computing power network is constructed by using the knowledge representation method. The scheduling tasks of computing power network is automatically executed by the proposed knowledge driven method based on the constructed knowledge graph. Different with the current scheduling technology of computing power network, the model will theoretically become more and more efficient and accurate with continuously addition of knowledge.