Nanjing Intelligent Computing Center: Based on Inspur distributed storage, accelerating the release of data potential

A Brief History of Mankind: From Animals to God by Yuval. Harari believes that the evolution and development of human beings from the Stone Age to the 21st century has gone through four stages, namely the cognitive revolution, the agricultural revolution, the integration and unification of human beings and the scientific revolution. At present, human beings are ushering in the second cognitive revolution, the era of artificial intelligence of intelligent computing!

From unmanned driving to precision medicine; from intelligent customer service to smart cities. Intelligent computing drives social change, and we have also stepped beyond weak artificial intelligence and stepped into the eve of the era of strong artificial intelligence. In this rapidly developing industry wave, the global industry has also transformed one after another, rushing to the beach to lay out the artificial intelligence innovation ecosystem.

According to IDC’s forecast, by 2024, China’s share of the global artificial intelligence market will reach 15.6%. China’s AI industry will become an important driver of global market growth.

Intelligent Computing Center: New economic momentum, new social infrastructure

On July 14, Nanjing Intelligent Computing Center was completed and put into operation in Nanjing Qilin Science and Technology Pioneer Park. The center cooperated with Inspur and Cambrian to create an industry-leading model of “One Center, One Base, and N Platforms”, that is, one center. Computing power support center, a PAAS ecological expansion base; N application platforms, focusing on supporting application innovation in the fields of technology finance, smart manufacturing, smart retail, smart medical care, and smart transportation. It provides five functional services such as intelligent computing software and hardware resource sharing and scheduling, cross-research support, big data application, unified security management and control, and industrial innovation agglomeration for multi-party users such as governments, enterprises and scientific research institutes.


Nanjing Intelligent Computing Center: Based on Inspur distributed storage, accelerating the release of data potential

According to the “2020-2021 China Artificial Intelligence Computing Power Development Evaluation Report” jointly released by Inspur and IDC, in my country’s 2020 AI computing power TOP 10, five cities in the Yangtze River Delta are on the list, and Nanjing is on the list. The completion of Nanjing Intelligent Computing Center will undoubtedly bring sufficient momentum to the transformation from digital economy to smart economy in the Yangtze River Delta region and even the whole country. By working with the government and enterprises to develop AI solutions for the smart city industry, relying on the platform to provide services such as innovation and entrepreneurship, education and scientific research, information sharing, forum competitions, etc., to build an ecological cooperation circle for intelligent computing, promote the integration of artificial intelligence technology with traditional industries, and further Promote the development of the artificial intelligence industry in Nanjing and the Yangtze River Delta region.

In terms of functional design, the solution of Nanjing Intelligent Computing Center follows the design principles and concepts of modularization, standardization, wide compatibility, independent safety and controllability, intelligence and efficiency, and effectively design and Configure the hardware structure of the system so that the hardware configuration architecture matches the requirements of scientific software, maximizes the capabilities of the hardware, improves computing efficiency, and meets the requirements of future development and system upgrades. In the construction, a new generation of artificial intelligence computing cluster architecture is adopted, combining versatility and specificity, short-term and long-term, while exploring the hot spots and frontiers of research in the fields of storage, computing, and algorithms to meet the needs of the intelligent computing business platform for informatization. While optimizing the power consumption and operation and maintenance costs of the facilities, it will jointly promote the development of the intelligent computing industry in Nanjing.

According to the “Nanjing City Action Plan for Building AI Industry Landmarks” issued by the Nanjing Municipal Party Committee and Municipal Government, by 2025, the scale of Nanjing’s AI core industry will reach 50 billion yuan, driving the scale of related industries to 500 billion yuan. It has developed into the first in the province, the top three in the country, and a globally influential demonstration city for artificial intelligence innovation applications.

Is there enough computing power to release the potential of data?

According to the “2020-2021 China Artificial Intelligence Computing Power Development Evaluation Report” released by IDC, among the three AI elements of data, algorithms, and computing power, computing power has become the driving force for AI industrialization and industrial AI development. key elements; at the same time, in recent years, the complexity of algorithmic models has shown an exponential growth trend. It can be seen that in the era of intelligence, computing power and data are becoming more and more important.

Have enough data? In such an era of Internet of Everything, for Nanjing, which is located in the Yangtze River Delta Economic Zone, the entire industry is rapidly entering the era of smart economic transformation. Hospitals, banks, financial centers, autonomous driving, smart transportation, online education, smart medical care, smart robots, Intelligent customer service, etc., the growth of data is exponential.

Have enough computing power? In July, the Nanjing Intelligent Computing Center was launched, with a computing capacity of 8 billion times per second (800P OpS). This is also the highest-performance intelligent computing center currently in operation in the Yangtze River Delta.

The overall system architecture of the artificial intelligence computing platform of Nanjing Intelligent Computing Center includes two layers, namely Inspur intelligent cluster hardware equipment and Cambrian artificial intelligence computing power platform. Inspur intelligent cluster hardware equipment provides basic hardware support for the entire intelligent computing center solution, including five parts: intelligent computing unit, data storage unit, network switching unit, support management unit, and information security unit, providing AI computing power and general computing power , high-performance switching network, all-flash storage, distributed storage, information network security and other basic support services.

The high-speed computing and storage network carries the bandwidth requirements for mutual access between intelligent computing servers and storage nodes, and connects the entire computing system and storage system. Redundant link design with storage nodes, 25Gb network rate access, access switches use 100Gb uplinks to connect to aggregation switches, and aggregation switches are interconnected with core switches through 100Gb redundant links.


Nanjing Intelligent Computing Center: Based on Inspur distributed storage, accelerating the release of data potential

Smart use of cloud storage, operational research of new data

The development of information technology has brought about the explosive growth of data. There is no doubt that we have fully entered the era of big data. Unstructured data of petabyte scale is becoming more and more common. How to effectively manage these data and further explore the value of data has been become an issue that must be taken seriously.

Nanjing Intelligent Computing Center builds a unified platform, sets up AI computing power resource pools and algorithm resource pools, builds a unified computing power and algorithm scheduling platform, and enriches algorithm resource pools through the operation of the Intelligent Computing Center and the construction of ecological chains to support the development of AI intelligent applications in various industries. training and inference. The operation process of Nanjing Intelligent Computing Center is extremely demanding on the performance and stability of the storage system. The storage system must carry tens of billions of small files to read and write, to meet the continuous and stable high concurrent access of thousands of clients on the deep learning platform, and to meet the needs of GPU in AI training. Or the concurrent access requirements of high-bandwidth and low-latency CPU computing clusters; with the linear growth of business deployment, the amount of data storage will also grow at PB level or even reach EB level, Nanjing Intelligent Computing Center also requires extreme storage device expansion capabilities. High; the complex AI model building process involves operations such as data collection, preprocessing, training, inference, and simulation. Different links correspond to storage services with different protocols, and a set of storage is required to support an intelligent computing center, which significantly shortens the AI ​​model generation time.

Inspur distributed storage AS13000 has become the most solid data storage base of Nanjing Intelligent Computing Center by virtue of its extremely fast performance, extreme stability, extreme capacity and multi-protocol integration leadership.

Extremely fast performance, large bandwidth of 100 GB, and millisecond-level latency. Based on the iTurbo intelligent engine, Inspur Storage AS13000 integrates intelligent IO balance, intelligent resource scheduling, intelligent metadata management, intelligent road cache acceleration and other system optimization modules to realize intelligent and efficient scheduling of key elements such as IO, paths, data blocks, and cores. The computing center provides millions of IOPS and large bandwidth of 100 GB, and realizes the retrieval and reading of tens of billions of small files in seconds.

Extremely stable, 99.9999% highly available, end-to-end guaranteed. Inspur Storage AS13000 builds multiple data protections at the application layer, data layer and system layer, and the operation of complex business systems is more stable. Among them, the industry-leading 32+2 large-scale erasure and deletion, the space utilization rate can reach 94%, ensuring high reliability and space for the system. The utilization rate is higher, and the 4T/H rapid reconstruction and security industry system switches the front-end business without perception, and runs continuously and stably.

Extreme capacity, EB-level capacity, trillion-level capacity. Inspur Storage AS13000 supports horizontal expansion to 5120 nodes, provides EB-level storage space, and linearly increases system performance and capacity. At the same time, the storage system supports multi-source zero-copy, soft-copy and other advanced functions to achieve the highest storage space utilization efficiency.

Minimalist architecture, multi-protocol interoperability, and high data usage efficiency. Inspur Storage AS13000 is the first in the industry to launch an “all-in-one” minimalist architecture. It supports four storage services of file, block, object and big data, enabling efficient interoperability of multiple protocols. One set of storage supports the entire business process and improves overall resource utilization. efficiency, and reduce maintenance and management costs for multiple sets of storage. A set of Inspur AS13000 storage supports the application of Nanjing Intelligent Computing Center, realizes multi-protocol and efficient interoperability in the AI ​​training process, and satisfies the fast reading of millions of files at the front end.

In the era of wisdom, by building a leading AI computing power infrastructure, the intelligent computing center will play a huge role in carrying intelligent technological innovation, promoting data open sharing, accelerating intelligent ecological construction, and driving the aggregation of intelligent industries. Inspur Storage will continue adhering to the “cloud storage and intelligent use, and operation of new data” to empower the intelligent computing center, accelerate innovation in the field of distributed storage, and release data value to drive intelligent transformation.

The Links:   SKM75GAL063D 6MBI25J-120

Author: Yoyokuo