Performance Analysis of Kubernetes Job Scheduling Model Based on Queuing Theory

Kubernetes is a distributed system infrastructure developed by Apache Foundation, which has become the key platform to process big data and has gotten more and more supports. For recognizing the huge potential of Kubernetes, while using the existing Kubernetes platform, more and more users also conduct performance analysis and test for Kubernetes platform. This paper analyzes the performance of job scheduling models under Kubernetes platform, using analytical method of queuing theory. There have been three existing scheduling algorithms under Kubernetes frame. They are FIFO, Capacity Scheduler and Fair Scheduler respectively. We establish appropriate model for the three algorithms based on queuing theory and analyze the performance. It is concluded that the performance of the system under M/M/S job scheduling model is better compared with existing M/M/1 job scheduling model.