Background

As the new era of intelligent networks arrives, communications networks will, on the one hand, fully absorb and use artificial intelligence to raise intelligence across network operations, service, and management, becoming users and beneficiaries of AI; on the other hand, they will greatly promote AI applications and industry, becoming drivers and enablers of AI. Artificial intelligence will be an important technology in the 5G era, or the intelligent-network era. The radio access network, core network, transport network, and other layers all depend on AI.

To meet the new network era, China Mobile is actively exploring and building the next-generation intelligent network. It has formed a relatively clear and complete development approach and has fully launched its artificial-intelligence layout and R&D practice.

Requirements

This China Mobile Research Institute project mainly supports China Mobile in accelerating a key AI technology platform and capabilities for intelligent operations and intelligent services, carrying out applied research and innovation on core algorithms in AI, and building a foundation AI platform. It is aimed at intelligent connectivity, intelligent decision-making, intelligent services, and other scenarios, providing basic services such as a deep-learning platform and core capabilities such as intelligent speech and face/image recognition, and carrying out AI applications in network intelligence, marketing, customer service, and other fields.

The GPU-server hardware must meet high reliability, high availability, and high scalability, and be a high-performance device that meets application requirements. It must support clustering and redundancy, active/standby or N+1 redundancy, load sharing, disk arrays, and other technologies to ensure highly reliable data and uninterrupted operation.

Solution

GPU servers in Sugon’s AI product series are a class of GPU servers aimed at medium-to-high power-density data centers and standard 19-inch racks, with flexible procurement and deployment. The 4U 8-GPU server used in this project is aimed at high-density computing environments that need high availability, high reliability, and high scalability, and is especially suited to deep-learning applications in artificial intelligence.

System-stability tests and GPU-card performance tests were conducted on GPU cards that met the requirements together with the GPU servers chosen for this project, and related test methods and reports were provided, strongly verifying product stability and high performance.