The next-generation cloud computing infrastructure achieves more efficient and flexible computing performance through unified management of heterogeneous computing (e.g., XPU) and a disaggregated data center architecture. XPU integrates different computing units such as CPUs, GPUs, and FPGAs, improving computational efficiency and resource utilization. The disaggregated architecture enables independent upgrades of computing, memory, and external storage resources, thereby enhancing the overall system reliability and scalability. Programmable distributed storage customizes storage policies and optimizes data management to meet diverse application needs.

Unified management of heterogeneous computing is an important technological trend. It aims to integrate different computing units, such as CPUs, GPUs, FPGAs, and various specialized acceleration chips, into a unified architecture. This approach enables more efficient and flexible computing performance, providing robust support for various complex computing scenarios and meeting the growing global demand for diverse types of computing power.

The disaggregated data center architecture enhances the flexibility and scalability of data center resource scheduling by separating computing resources from memory and external storage. This architecture allows resources to be upgraded independently.