Design of power information intelligent scheduling model based on Kubernetes

In order to solve the problems of complex service configuration, unbalanced utilization of computing resources and high operation and maintenance cost in the current artificial intelligence scheduling system, this paper proposes a solution of container scheduling and service management based on Kubernetes. According to the demand of intelligent scheduling engine platform, the modules of the platform are designed from the aspects of function realization and platform architecture. At the same time, to solve the problem that Kubernetes cannot sense GPU resources, device plugin is introduced to collect GPU information on each node in the cluster and report it to the scheduler. In addition, considering that the optimization algorithm in Kubernetes scheduling strategy only considers the resource utilization and balance of nodes, a optimization algorithm based on Pearson correlation coefficient is proposed. Pod scheduling is determined by calculating the complementarity between container resource demand and node resource utilization, so as to ensure the resource balance of nodes after scheduling.