GPU Operator Certified

Integrates components such as GPU-device (device discovery and health monitoring), driver-manager (driver deployment and upgrade), and container-runtime (container runtime integration), achieving full-process automated management from device discovery → driver installation → resource scheduling → status monitoring.

• Driver deployment strategy: Supports PreferCloud (image driver priority), PreferHost (host driver priority), etc., suitable for various usage scenarios
• Gradual upgrade strategy: Through configuring cluster driver batch upgrades (supports controlling the number of parallel nodes and defining failure rollback strategies), ensuring business continuity
• Device binding strategy: Flexible and powerful driver support, supports binding VFIO-PCI drivers by node or GPU serial number, meeting the requirements of virtualization scenarios

• Supports mainstream Linux OS systems, especially Red Hat Enterprise Linux and RHCOS
• Compatible with Kubernetes 1.23+ versions, supports container runtimes such as containerd, Docker, and cri-o

Built-in data-exporter component, supports Prometheus metric export (such as GPU utilization, temperature, etc.), flexibly exposes monitoring interfaces, seamlessly integrates with an enterprise-level monitoring system.