Amid fierce competition for AI development among companies, IT organizations are under pressure to provide an AI development environment in a timely manner. You are in the process of combining multiple open sources and deploying them with containers and Kubernetes.
In particular, open source Kubernetes increases the load because software is difficult to use. This is why ‘enterprise-grade Kubernetes’ from solution companies that lower barriers to entry are attracting attention.
A typical Kubernetes enterprise platform is the Red Hat OpenShift platform. Kubernetes adds advanced management and security features to meet the needs of businesses. Various convenient functions greatly ease the burden on operators and developers.
Red Hat OpenShift makes it easy to create a machine learning environment. With the application suite provided by OpenShift, create a machine learning development environment in minutes. Provision of separate environments for each of the multiple users is also simplified.
Supported by OpenNaru, a specialized Red Hat partner, this document directly built the systems required for a machine learning development environment in 10 minutes with Red Hat OpenShift and experimented with a simple machine learning environment with a Jupyter laptop.
▲ Preparation: Red Hat OpenShift 4.6 (based system infrastructure is ready)
First, access the Red Hat OpenShift Container Platform console in a web browser. From the administration menu, go to the operator ‘Operator Hub’. Operator Hub is a bundle of projects built in advance by Red Hat, partners, and communities. You can search and select the items you want in various fields, such as machine learning, application runtime, big data, and databases.
Enter ‘Open Data Hub Operator’ in the search bar and check the package information. Choose between automatic installation or manual installation to start the installation process. The installation work takes a few minutes.
When all the installation work is completed, you can check the related system information. If you go to the ‘Root’ menu on the left sidebar, you can check the network details of the services. The ‘Workload’ menu in the sidebar shows information for each node.
Copy the ‘Jupyter Hub’ node URL link from ‘Workload’ and paste it into your browser’s address bar and launch it. When the Jupyter Notebook login screen appears, you can log in using the authentication process. You can check out the available laptop images and choose the one you want. You can choose the container size and the amount of GPU.
After logging in, the Jupyter notebook environment appears. You can load any dataset you want and run machine learning jobs in the editor. The data sample used for the demonstration is data from the British Titanic, which sank in 1912.
Red Hat OpenShift deploys applications per container. If multiple users want to use a separate Jupyter notebook, they can clone and map an existing container.
Related Posts
Kim Young-joo, Open Naru Solution Business Team Team Leader, said: “When operating with open source, the issues are performance, stability and security. They check everything down to the last name, provide it, and take responsibility for it. . “
He emphasized that “OpenShift provides convenience to business users and reduces infrastructure concerns by changing the operation of the data center from device-oriented to application-oriented.”