Xinhua 3: In the AI era, enterprises need to have the ability to "adapt on demand"

Global Network Technology Reporter Lin DiWith the continuous launch of large models, the wave of AI industry is becoming increasingly fierce. In today's rapidly changing market environment, as an important infrastructure for enterprises, the network needs to have the ability to adapt on demand to support business iteration and model innovation

Global Network Technology Reporter Lin DiWith the continuous launch of large models, the wave of AI industry is becoming increasingly fierce. In today's rapidly changing market environment, as an important infrastructure for enterprises, the network needs to have the ability to adapt on demand to support business iteration and model innovation.

How will future AI trends and connectivity technologies evolve? How can enterprises empower their own product technology development with AI? The reporter interviewed Zeng Fugui, Vice President of Xinhua Third Group and President of Network Product Line, and Cheng Zhen, General Manager of Network Product Line System Planning and Solutions Department of Xinhua Third Group, on these issues.

Regarding the development trend of the entire industry brought by AI, Zeng Fugui shared: "In the AI era, there are some trends in the network. For example, rapid model innovation and rapid business iteration mean that the Digital transformation of all walks of life is accelerating. At the same time, it has also expanded some new industrial tracks, and upgraded the old industries. Today, AI has been in full bloom, and there are more and more commercial scenarios. In short, AI has brought a tremendous change to our lives, industrial upgrading, and improving production efficiency

Zeng Fugui believes that the rise of large models has directly led to explosive growth in computing represented by GPU and DPU. In addition, the large model also presents some new challenges and requirements for high-speed network connectivity. For the internet, this is a new change and also a new opportunity. Specifically, the network plays two roles as a connection, one is to efficiently connect the entire computing power, and the other also needs to be connected through the network during the use of computing power.

For Xinhua San, in addition to providing diversified computing resources, there are also abundant network devices. In this era context, Xinhua San believes that there are still three directions for the evolution of network devices: high bandwidth, low power consumption, and low latency.

It is reported that Xinhua San has made various explorations on the network, both in software and hardware. For example, it utilizes CPU silicon light, liquid cooling technology, and built-in AI algorithms to find a comprehensive solution between high-performance and green.

Zeng Fugui further pointed out that Xinhua has deeply laid out the entire industry chain of "cloud network computing storage end", continuously improving the level of digitalization and intelligence empowerment through these aspects, and providing a solid foundation for the innovative development of the big model field.

The combination of network, storage, and computing was originally relatively fragmented and developing separately. However, with the deepening of AI applications, we have integrated network, storage, and computing, and there are three points that need to be paid attention to throughout the training process of the large model. "He told reporters, first, the network has a large bandwidth. In the entire AI large model training, the consumption of computing resources is very high, which also puts higher demands on network bandwidth. Second, for the requirements of data Transmission delay, the data Transmission delay can be further reduced after the combination of network, storage and calculation. Thirdly, by endowing existing networks with computing power and utilizing specific algorithms, data exchange between GPUs can be reduced, thereby further improving the computational efficiency of the entire AI system and AI model. So the integration of these three brings a significant benefit to the entire large model calculation, which is efficiency improvement.

It is reported that in response to this, Xinhua San also provides a full stack solution for intelligent computing data centers, effectively integrating network, storage, and computing, thereby improving the efficiency of the entire big data model training.

As the industry knows, in terms of computing power acceleration, AI has brought significant changes to various industries, while also empowering the network itself. This requires the entire network to be an open architecture with programmable capabilities, and the network can be upgraded and evolved based on AI, making it more intelligent.

It is worth noting that with the changes in technology, business is becoming increasingly rich. Xinhua San proposed the "application driven network", which is not only connected, but also intelligent and can evolve itself. Xinhua San can regard the network as an important part of building the entire computing infrastructure, making it a building element of the entire system and a service.

For example, when a high traffic application appears, the network will automatically allocate more bandwidth and resources to ensure that application performance is not affected. At the same time, network management will also automatically adjust according to the needs of the application, such as implementing stricter security controls on highly sensitive data. The significance of network autonomy lies in the fact that networks can automatically monitor and handle problems that arise, such as identifying and isolating network attack behaviors. Such a network can be more reliable and secure, and also more adaptable to high loads and constantly changing application requirements.

As is well known, as AI capabilities grow, security issues become increasingly important. Cheng Zhen explained that AI requires data centralization and model training. On the other hand, data is also a valuable asset for enterprises, so protecting data is indeed a major challenge in the computing power era.

Regarding this, Cheng Zhen pointed out that Xinhua San provides network security protection through the following methods:

Firstly, adhere to technological innovation and the concept of safety. Last year, Xinhuasan released the active security 3.0 strategy. The core idea is to change the security defense mode from passive defense to Active Defense. When building a security system, instead of conducting traffic control at the border, we combine the results of cloud, network, and end monitoring to see where security risks exist.

Secondly, Xinhua San has done a lot of work in AI empowerment security. For example, by introducing more AI capabilities into security platforms, AI can empower security detection, intelligently detect and quickly defend attack behaviors, and achieve rapid response effects.

Thirdly, in terms of forward-looking defense, unlike before patching or upgrading after problems were exposed, Xinhua San conducts trend judgment and early warning of risks.

Fourthly, tracing the source of threats and establishing trend analysis, Xinhua Three will conduct model analysis and source tracing of attack sources. Through cloud network integration, intelligent judgment of the source is carried out to accurately block the attack source, especially through linkage between cloud and network.

In the era of AI and big models, the explosive growth of computing power has brought about changes and innovation in technology. Network devices need to cope with the current explosive growth trend of computing power and meet the high bandwidth and large capacity computing needs of data centers. At the same time, it is also necessary to meet the requirements of energy conservation and consumption reduction in the construction of green data centers.

Disclaimer: The content of this article is sourced from the internet. The copyright of the text, images, and other materials belongs to the original author. The platform reprints the materials for the purpose of conveying more information. The content of the article is for reference and learning only, and should not be used for commercial purposes. If it infringes on your legitimate rights and interests, please contact us promptly and we will handle it as soon as possible! We respect copyright and are committed to protecting it. Thank you for sharing.(Email:[email protected])