Goal

This feature integrates NVIDIA GPUs into bare metal servers, enabling the execution of AI workloads in confidential containers. It's designed to protect the confidentiality and integrity of sensitive data, making it feasible for highly regulated customers to use on-premises bare metal infrastructure for their AI needs while adhering to strict compliance standards.

Expected user outcomes

The observable functionality that the user now has as a result of receiving this feature.

1. Enhanced AI processing: Confidential NVIDIA GPUs boost the power for AI tasks, enabling more complex computations.

2. Secure data handling: Encrypted GPU memory offers heightened security for sensitive data processing.

3. Minimized security risks: Removing the GPU from the trusted computing base reduces potential vulnerabilities.

4. Compliance in sensitive sectors: Additional security features assist in meeting strict industry regulations.

5. Expanded AI possibilities: The integration allows new AI applications in data-sensitive environments.

Acceptance Criteria

1. Seamless GPU Integration: The feature must integrate perfectly with confidential NVIDIA GPUs in bare metal servers, ensuring no compatibility issues.

2. Robust Data Encryption: Implement industry-standard encryption protocols for GPU memory to secure sensitive data effectively.

3. Effective TCB Reduction: Achieve a significant reduction in the trusted computing base, especially in relation to GPU integration.

Questions to Answer (Optional):

Installation of Nvidia GPU drivers in the Kata VM image for bare metal. Are these drivers part of RHEL or available in a public repository?