Inspur Information Supports the Advancement of AI Services with South Korea\'s Leading Mobile Communications Operator

Background introduction:

  • Customer Introduction:

A leading mobile communications operator in South Korea is committed to becoming an "artificial intelligence and digital infrastructure services company". By strengthening its competitiveness in three key business areas: fixed and wireless communications, AI services, and digital infrastructure services, it anticipates an annual revenue target of 22 trillion KRW by 2025.

  • Project Background:

The vigorous development of AI has brought new opportunities for telecommunication networks. The ability to handle both data and information is an important capability for operators in the digital transformation era. AI, with its powerful data analysis and information extraction capabilities, helps operators transform data into information. The industry hopes to help solve various efficiency and capacity problems currently encountered by communication networks by introducing AI, and in turn provide vastly improved integrated digital and information services. In 2016, this particular customer officially entered the AI market and launched the first Korean digital services assistant. Following that, the customer launched an AI service integrating all of the company's products and services, providing personalized services according to customer preferences.

Challenges:

The application and business development of AI in telecommunication networks requires upgrades across the board, particularly in data usage and computing power. An AI services centered business model faces the following challenges:

1. AI services need a powerful AI computing platform

Based on the large language model GPT-3, the customer independently developed a Korean large-scale natural language processing model, and launched an AI service that can conduct Korean dialogue with customers. Based on large-scale natural language processing and emotion analysis technology, the service provides customers with daily conversation, information acquisition, content broadcasts, and destination navigation services. The powerful data processing capabilities and model training required by this service needs to be supported by a powerful AI computing platform.

2. High complexity of data center management

With the expansion of the customer’s business scope in the field of AI services, a large amount of manpower needs to be invested in the batch deployment and daily maintenance of hardware equipment, necessitating the urgent need to improve deployment and maintenance efficiency. In addition, facing a huge scale of users, business continuity poses more challenges to rapid fault location and maintenance efficiency of data center equipment.

Solution introduction:

In order to better support the AI computing power requirements of the customer’s Korean language model, Inspur has provided the leading NF5488A5 AI server along with cluster system design and optimization services. The ISPIM server management platform is provided to simplify daily operation and maintenance management.

Detailed Solution

Multiple GPU clusters based on Inspur NF5488A5 AI servers were deployed. Inspur NF5488A5 is equipped with 2x AMD EPYC CPUs and 8x NVDIA A100 SXM4 GPUs, providing strong AI computing power. The NVMe SSDs and CPUs are connected directly to further reduce I/O latency. A single machine can provide up to 20PetaOPs of computing performance, and its floating point computing power can reach 5Petaflops. Additionally, Inspur provides diversified value-added services, such as cluster system design, cluster performance optimization, application software optimization and other professional services to help customers effectively use infrastructure resources while accelerating the launch of the AI development environment. In addition, Inspur provides the ISPIM server management platform for full life cycle operation and maintenance management. This platform can realize the integrated management of multiple equipment in the data center, vastly improving maintenance efficiency.

Customer Benefits

The AI solution architecture built by Inspur maximizes the overall AI architecture computing power of the customer, meeting the computing power requirements for efficient operation of large-scale natural language models and other applications, helps R&D personnel improve development efficiency, improves customer operation and maintenance efficiency, and provides a solid hardware foundation for the customer’s AI business transformation.

Recommend Products

Related Solutions