KubeDay Japan 2024

Managing kernel modules for AI/ML workloads in Kubernetes within a production environment has been a cumbersome manual process. This session introduces the Kernel Module Management operator which addresses challenges posed by AI/ML workloads, edge computing, and software-defined storage solutions. The presentation will delve into the complexity of KMM, showcasing its ability to streamline the lifecycle management of kernel modules at scale. Attendees will also discover how to simplify the traditionally complex task of customizing nodes in Kubernetes, thereby reducing cost. The operator's capacity to build OCI images and sign kernel modules in-cluster ensures compatibility with Secure Boot, adding an extra layer of security. A live demo will show the practical application of KMM, highlighting its features and capabilities. The session will conclude with a production example, demonstrating how KMM significantly simplifies the lifecycle management of GPU drivers in Kubernetes at scale.