With the rise of the AI era, GPUs are being widely used in various applications, including deep learning, scientific research, medical imaging, video rendering, engineering manufacturing, remote sensing mapping, financial analysis, and weather forecasting. However, due to the high cost, customers often choose to optimize GPU utilization through effective management instead of purchasing a large number of GPUs. Key challenges faced by users during GPU utilization include:

1. Inefficient GPU Server Management

Enterprises often purchase a large number of GPU servers without a unified O&M management platform, lacking cloud platform support for GPU server application, allocation, and deallocation.

2. Lack of Visual Management for GPU Resources

A lack of visual management tools exists in GPU resource management scenarios. Instead, command-line interaction is used for GPU resource management.

3. Low GPU Resource Utilization

Traditional graphics workstation architectures are siloed, with limited sharing capabilities, resulting in inefficient scheduling and allocation of GPU resources and low utilization of device resources.

4. Complex Management and Maintenance

Managing and performing O&M on graphics workstations with multiple heterogeneous devices can be challenging. Different design software requires different GPU resources, and the underlying architecture lacks flexibility, making it difficult to adapt efficiently. In addition, system expansion suffers from complexity and limited scalability.