During the China Computing Power Conference on September 29, the '2024 Intelligent Computing Era Datacenter Technology Leadership and Transformation Forum', jointly organized by the China Academy of Information and Communications Technology (referred to as 'CAICT') and Chindata Group (referred to as 'Chindata'), was successfully held. The forum, with the theme of 'Creating a New Ecosystem for Integrated Industry, Building a Smart Computing Future', brought together experts from multiple industry, academia, and research institutions such as CAICT, China Telecom, Intel (China), Baidu, Alibaba Cloud, Inspur Electronic Information Industry, Seagate Technology, Chindata, to deeply focus on the development direction of intelligent computing, IT equipment, and other cutting-edge trends in the era of artificial intelligence, and collectively discuss the challenges and breakthrough paths faced by data centers in the face of new trends.

At the event, in response to the trends of high-density, large-scale, and heterogeneous computing power development in the intelligent computing era data centers, Chindata upgraded and released the 'Intelligent Computing Center Full-Stack Solution 2.0', to systematically innovate and fully support the computing power demand in the intelligent computing era. In response to the trend of liquid cooling in data centers, Chindata, together with its ecological partners, jointly launched the establishment of a series of industry standards for lifecycle quality control of the cold plate liquid cooling system, set up an expert group, and accelerated the standardization and normalization process of data center cold plate liquid cooling systems.

Upgrade and release of the 'Intelligent Computing Center Full-Stack Solution 2.0' targeting AI scenarios

Currently, with the continuous increase in GPU and server power consumption, the computing power cluster is evolving towards a scale of over a hundred megawatts, posing new challenges to traditional data center operations. As a leading operator of computing infrastructure solutions, Chindata maintains a forward-looking approach to computing power and technological development, creating the 'Panshi' modular data center framework and continuously iterating to tackle the five core challenges of high computing power, ultra-large scale, high elasticity, high efficiency, and greater intelligence.

In the 'Panshi' framework, Chindata, together with upstream and downstream ecological partners, jointly released the 'Intelligent Computing Center Full-Stack Solution 2.0' at this forum. The solution focuses on four major product segments: high-elasticity building model, 'Xuan Iron' power supply system architecture, 'Xuan Bing' cooling architecture, and intelligent operation and maintenance. Compared to previous versions, the newly released solution is more advanced in technology, with higher levels of modularization and integration, fully meeting the demand for larger computations in the era of large models and General Artificial Intelligence (GAI), and supporting agile delivery and elastic scaling of customer businesses more quickly and with higher quality.

Among them, the modular architecture of the high-elastic building model can be flexibly configured and prefabricated in stages, providing strong support for business elastic deployment, while being compatible with different computing power scenarios of CPUs and GPUs, increasing the IT output per unit area. This solution effectively addresses challenges such as the rapid growth of high-density requirements and heterogeneous computing power, pioneering a new mode of rapid delivery for data centers.

The modular "Mystery Iron" minimal power supply architecture integrates more than 10 links of traditional power distribution into six electromechanical modules: substations, diesel generators, medium voltage, low voltage, batteries, and IT. The modules can be deployed and expanded in phases to ensure high-quality delivery and fast delivery pace.

The "Mystery Ice" cooling architecture includes modules such as air cooling, liquid cooling, air-liquid compatibility, etc., collectively solving the cooling requirements for high-density calculations and addressing the cooling challenges for power densities ranging from 8kW to 150kW in cabinets. Different cooling technologies and products can be flexibly combined to achieve the best cooling solution, breaking through limitations of buildings, climates, and regions.

Intelligent operation and maintenance rely on the in-house monitoring platform of Chindata + IDC operation management platform to drive frontline operation and maintenance into a more standardized and intelligent era of intelligent data center management. Simultaneously, in collaboration with AI companies, the development of AI solutions optimizes the control of cooling systems.

CTO Zhang Binghua of Chindata Group stated: "The AI wave profoundly impacts the technological transformation path of data centers. Chindata has always been committed to forward-looking innovation, conducting top-level innovation throughout the entire business cycle. This upgrade significantly enhances technological capabilities and focuses on innovation towards modularity, distributed systems, hyperconvergence, and prefabrication, enhancing the elastic capabilities to support different computing scenarios, thereby solidly establishing the computing power foundation for the intelligent computing era, achieving efficient transformation from power to computing power."

Launch of the Four Major Standards Accelerates the Standardization Process of Cold Plate Liquid Cooling

In the era of intelligent computing, traditional air cooling technology encounters cooling efficiency bottlenecks. With its outstanding heat dissipation performance, cold plate liquid cooling technology has become the preferred cooling solution for large-scale intelligent data centers. However, the past custom focus of cold plate liquid cooling systems has led to significant quality differences, increasing the difficulty in design and deployment.

To promote the standardization and compatibility of cold plate liquid cooling systems, at the forum, the "Series of Industry Standards for Quality Control Specifications for the Entire Life Cycle of Cold Plate Liquid Cooling Systems" (referred to as "Series Standards") was officially launched for project establishment, with the expert group established concurrently.

The series of standards will be led by China Academy of Information and Communications Technology and Chindata, jointly developed with dozens of leading enterprises in the industry. The series of standards include the "Standard for Quality Control of Cold Plate Liquid Cooling System Design", "Standard for Quality Control and Detection Methods of Datacenter Cold Plate Liquid Cooling System Materials", "Standard for Quality Control of Cold Plate Liquid Cooling System Testing", and "Standard for Quality Control of Cold Plate Liquid Cooling System Operation", aiming to provide practical guidance on quality control for various stages of the cold plate liquid cooling lifecycle, promote the construction of upstream and downstream ecology, and create a positive cycle for faster technological innovation and larger scale commercial use.

Before the series of standards are launched, China Academy of Information and Communications Technology, Chindata, and other units jointly compiled the White Paper on Quality Control Standards for Datacenter Cold Plate Liquid Cooling Systems, based on Chindata's deployment practices of multi-megawatt level cold plate liquid cooling, operational experiences from over 30 companies, and international related standards. It systematically proposes for the first time a quality management and guidance system covering the entire lifecycle of cold plate liquid cooling. This white paper will provide important references for the formulation of the series of standards.

Zhang Binghua said that industry standards are crucial for unifying technical standards and regulating industry ecology. We will strive to gather the largest consensus in the industry, explore the optimal balance between stringent indicators and large-scale deployment, establish mainstream standards widely recognized in the industry, thereby accelerating the technological advancement of cold plate liquid cooling and contributing to the green and sustainable development of the data center industry in the era of intelligent computing.