Improving the Efficiency of Container Image Deployment through a Cluster Container Registry

Large-scale machine learning models and data analysis applications use container images to process vast amounts of data, and there is a problem of lengthening the distribution time of such images Longer image distribution time slows down the development and distribution cycle, which lowers productivity and can reduce the performance and flexibility of the entire system. In order to solve this problem, we propose ways to improve container image management and distribution efficiency. We focus on how to reduce time and traffic consumption compared to image downloading through the cloud by utilizing cluster container registry (CCR) when distributing images. Experiments were conducted in a Kubernetes, and the efficiency of container image management and distribution process was explored.