logo

Operating CERN SaaS at Scale with Operators

2023-04-19

Authors:   Michael Hrivnak, Rajula Vineet Reddy, Francisco Barros, Varsha Prasad Narsing


Summary

CERN uses the operator pattern to automate and scale delivery of CMS websites, balancing reusability and open source principles against integration with CERN’s specific compute environment and existing infrastructure services.
  • CERN operates 1000+ CMS websites as a SaaS running on Kubernetes
  • The small team used the operator pattern to automate and scale delivery of CMS websites
  • Balancing reusability and open source principles against integration with CERN’s specific compute environment and existing infrastructure services
  • Operator SDK, its best practices, and things to avoid when developing an operator from scratch
  • How Kubernetes enables isolation, multi-tenancy, and resource sharing
  • Automated maintenance and monitoring
CERN hosts a lot of websites and manages content management systems, specifically Drupal instances. They wanted each instance to have a life cycle to be autonomous and isolated, and to enforce all the dependent resources. With the operator pattern, they were able to create and delete sites on the fly, reduce inconsistencies and errors with manual creations, and scale up easily. They were able to standardize configuration and the flavor of quality of service.

Abstract

CERN operates 1000+ CMS websites as a SaaS running on Kubernetes. This expert panel of end users from CERN and engineers from Operator Framework will discuss lessons from CERN’s drupal operator including: - How a very small team used the operator pattern to automate and scale delivery of CMS websites. - How they balanced reusability and open source principles against integration with CERN’s specific compute environment and existing infrastructure services. - Operator SDK, its best practices, and things to avoid when developing an operator from scratch. - How Kubernetes enables isolation, multi-tenancy, and resource sharing. - Automated maintenance and monitoring. Following the initial discussion of approximately 20 minutes, we will leave substantial time for Q&A. The target audience is anyone who is planning to build a SaaS on Kubernetes or operate many instances of an application.

Materials:

Post a comment

Related work