On the afternoon of June 14th, Sina Technology reported that Huawei Cloud's East China (Wuhu) data center was officially launched, marking the completion of the national storage and computing network layout of Huawei Cloud. The hub node of the Yangtze River Delta for both computing and storage can provide super large-scale computing services to the East China and central regions, which will effectively promote the development of the AI industry in the relevant regions.
Huawei's vice chairman and rotating chairman Xu Zhijun said that Huawei Cloud's East China (Wuhu) data center is an important part of Huawei's cloud blueprint. It can help enterprises use cloud computing power, use it as needed, and keep up with the latest technology. Enterprises do not need to build data centers and can focus on their own business development.
Zhang Ping'an, executive director of Huawei and CEO of Huawei Cloud, said that Huawei Cloud has redefined cloud infrastructure with systematic innovation. In the future, it will promote innovative applications of various shapes through the global storage and computing network and the Ascending AI Cloud service, and accelerate the intelligent transformation process of thousands of industries and enterprises.
Zhang Ping'an introduced that the Wuhu Data Center is planned to cover an area of about 6,000 acres, with a planned 3 million servers. Adhering to Huawei Cloud's rich experience in energy saving, AI clusters and liquid-cooled data centers, the Wuhu data center builds technology advantages such as green, low carbon, secure and reliable.
At the same time, in order to meet the demand for super large-scale computing power, the Wuhu Data Center has adopted Huawei Cloud's newly developed multi-power equivalent pooling architecture for the first time: through the distributed engine architecture, CPU, NPU, GPU, memory and other diverse resources are uniformly abstracted and pooled, evolving from a single computing power architecture to a matrix computing power architecture, which increases computing power by 50 times; through the high-bandwidth Scale Up network, resources are equitably connected, breaking the performance and cluster linear bottlenecks of single computing powers, and effectively improving the training efficiency of large models; in addition, through the Yao Light Intelligent Cloud Brain, various pooled resources can be uniformly scheduled and managed, achieving flexible and on-demand combinations of the best computing power ratio for models ranging from billions to trillions of parameters.
Wuhu has good conditions for large-scale data center clusters, and fiber networks can directly reach hot cities in East China and central China with network latencies within 10ms, with Shanghai and Hangzhou at around 5ms and Nanjing at less than 2ms. It can meet the needs of businesses that are not sensitive to latency, as well as those with extremely high requirements for latency, such as AI inference, e-commerce, games, etc. It is the preferred choice for enterprises with high computing resource consumption such as automobiles, industry, biomedical, and high technology in the Yangtze River Delta.
"Relying on the full-stack AI capabilities built by Huawei Cloud, through innovative technologies such as global storage and computing network, Ascending AI Cloud service, and Pang Universe's large models, we will inject intelligent vitality into creative development of thousands of industries and accelerate the cultivation of innovative and competitive new production forces in China's intelligent industry," said Zhang Ping'an.