20 data centers in the provinces of Guangdong, Sichuan and Guizhou have become the first data centers to connect to C2NET, an ambitious project to consolidate and distribute computing power in China.
C2NET’s computational capacity is already over three exaflops, meaning it can perform three quintillion (10th power to 18th power) calculations per second. This far exceeds the capabilities of the most productive Frontier supercomputer in the world – 1.2 exaflops.
The demand for high-performance data centers, thanks to the growing popularity of artificial intelligence, is growing rapidly around the world, and China is thus creating the necessary infrastructure for both AI research and technology development. of the computer industry in a context of restricted access to advanced technologies. chips from the American company NVIDIA introduced from the United States.
The C2NET project is part of the “Eastern Data and Western Computing” plan to address regional disparities in China’s digital resources. According to the plan, data collected in industrial and densely populated areas in eastern China will be sent to data centers in western regions of the country with lower population density, land and electricity costs. These regions have significant renewable sources of wind and solar energy, and there are plans to power new data centers from these.
Data centers in the east will primarily be used for real-time information processing – for time-sensitive sectors such as e-commerce, video training and the gaming industry. Data that can be processed with a delay will be sent to Western data centers. In February 2023, the country’s first computing power exchange opened in China, which provides pooling, distribution and access to computing resources for businesses and organizations. Major companies such as Huawei, ZTE, Alibaba and China Telecom have already joined the exchange.
“Computing resources will become a kind of commodity, like electricity. You can pay for their use when you need them, no matter where the source is,” said Gao Wen, director of the Shenzhen State Laboratory, who leads the C2NET project.
He also noted that computing resources are essential for the development of large language models (LLM), the technology used to train in-demand chatbots like ChatGPT.
To date, Chinese companies have launched at least 79 LLMs with over a billion parameters each.
According to the TOP-500 international ranking of supercomputers, China and the United States now occupy the first places in terms of the number of supercomputer systems. In June 2023, the United States jumped to the top spot, boosting its supercomputer count from 126 systems in November’s list to 150 in June. China fell from first to second place in June, reducing its representation in the TOP-500 from 162 computer clusters to 134. Russia was represented by seven supercomputers in the November and May lists.
The shortage of computing power in Russia, particularly in the context of sanctions, the withdrawal of Western suppliers and the inability to work with American cloud providers such as AWS, Microsoft Azure and Google Cloud, has put companies , especially small and medium, in a difficult position. The business model of small businesses and startups does not involve investing in their own IT infrastructure and makes them dependent on external data centers and cloud providers.
Thus, small and medium-sized enterprises have found themselves in a less favorable position compared to large enterprises that have their own data centers. This is indirectly confirmed by the results of a recent study conducted by the experts of the international award in the field of digital technologies “Skillful Intelligence”. So, speaking about the reasons for the failures of Russian startups, the majority of survey participants – 36.8%, called the problem of monopoly in the market. According to them, young companies fail to grow because it is difficult for them to compete with large companies. This competition also includes a competition for access to computer resources.
Considering China’s huge resources in this field and the creation of a national “super cloud” C2NET, “RG” asked experts to comment on the possibility of using Chinese computing power by Russian companies.
CorpSoft24 CEO Konstantin Renzyaev assessed this business model very cautiously: “There are several points to consider. First, is it fair for Russian companies to fall into the arms of Chinese infrastructure? We have already worked in European data centers. It was convenient, inexpensive, but we all know “How it ended. In addition, in the case of leasing infrastructure located in China, one must take into account the delays in the communication channels from China to the European part of Russia, where the main consumers are located. ”
According to Renzyaev, potential language issues with Chinese tech support and exchange rate differences need to be considered. “I don’t think the services of the Chinese infrastructure, despite its size, will be in demand,” said the boss of CorpSoft24.
Sergey Zinkevich, Business Development Manager of CROC Cloud Services, also believes that despite the obvious advantages of Chinese capabilities, such as scale and performance, there are many hidden pitfalls that negate the benefits of their use by Russian companies: “For example, training a neural network requires a huge data network that needs to be transferred to Chinese data centers. Turns out the company drives traffic to China for a week, trains the neural network for a week, brings traffic back for a week. in general, the task will take three weeks, during which you will have to pay for the capacity. one week cloud rental is enough for the same task. Therefore, even if the cost of the Russian cloud is higher, the use of Chinese capacities may prove to be less economical, taking into account all the associated costs and risks.
Additionally, according to Zinkevich, there may be issues with settlements, payment, the Chinese firewall, which can block some traffic, and other nuances of Chinese law that Russian companies are unaware of: “Therefore, before making such a choice, you should seriously consider all possible options and weigh their pros and cons.”
Read the Latest Science and Technology News Today on The Eastern Herald.
