Home > Technology peripherals > AI > body text

Trend Technology: Resource pooling releases the value of AI computing power

WBOY
Release: 2023-10-17 11:41:05
forward
652 people have browsed it

"According to statistics, the average utilization of GPU does not exceed 30%, which will cause a huge waste of computing resources. We use software-defined methods to usually increase the utilization of users' GPU by 3-8 times, or even to 10 Times."

This is the latest data provided by Trend Technology, a computing power pooling software company, citing industry report data and combining its own corporate best practice experience. Through the self-developed OrionX AI computing resource pooling software, Trend Technology can help customers build a data center-level AI computing resource pool. Customers can use the AI ​​computing power of any server in the data center without modifying the application, and By greatly optimizing application deployment and improving user resource utilization, it brings a new computing power supply model to the AI ​​industry.

Trend Technology: Resource pooling releases the value of AI computing power

Redefine AI computing resources

In the second photosynthetic organization AI solution competition, Trend Technology created the "Xinchuang full-stack independent controllable heterogeneous AI computing power resource pool" based on the adaptation and optimization of Haiguang hardware product base (hereinafter (referred to as "AI Computing Resource Pool") solution won the first prize in this competition.

Wen Yuxiang, Ecological Operation Director of Trend Technology, introduced that the AI ​​computing power resource pool solution mainly addresses the current extensive use of heterogeneous computing power and the difficulty in unified management and scheduling of heterogeneous computing power resources across manufacturers. The solution pools and manages physical heterogeneous computing resources to achieve on-demand scheduling of heterogeneous computing resources and refined management of cross-vendor heterogeneous computing resources.

Trend Technology: Resource pooling releases the value of AI computing power

Project Architecture Diagram

The infrastructure layer in the solution can achieve two major leaps based on the performance advantages of Haiguang servers; at the operating system layer, the AI ​​computing resource pool is based on the Galaxy Kirin server operating system, with the help of the advantages of Haiguang CPU and DCU platforms. Kernel security, RAS features, I/O performance, virtualization and domestic hardware, driver support, etc. are further optimized and enhanced; at the virtualization layer, OrionX abstracts AI computing resources into virtual resources, and ultimately realizes AI computing resources software definition.

“Currently, resources are directly connected and exclusive in application deployment, resulting in problems such as waste of computing resources and inflexible use. The cost burden on computing power users continues to increase, which has hindered the further development of AI technology to a certain extent. We hope that computing power, which in the past seemed like a piece of cake, can truly become a resource that benefits the public, just like water and electricity, which can be used as needed, flexible, convenient, cheap and efficient," Wen Yuxiang said.

Let the ecological flywheel speed up

Recently, an article titled "Research on the Application of Artificial Intelligence GPU Computing Resource Pooling" signed by the CIO of Hua Xia Bank was published in the journal "Financial Electronicization" managed by the People's Bank of China. This article discusses in detail the application value of heterogeneous computing power resource pooling software in the field of information innovation, and further verifies the feasibility of OrionX heterogeneous computing power pooling software solution of OrionX. It also evaluates the performance of domestic hardware such as Haiguang accelerator cards. affirmed.

"The construction of computing resource pool in the field of information innovation needs to be jointly promoted by domestic software and hardware manufacturers. Haiguang is in the leading echelon of the domestic chip industry and provides us with great support." Wen Yuxiang said that through cooperation with Haiguang , Trend Technology has realized the diversification of heterogeneous computing power of pooled software, further meeting the application needs of the Xinchuang market.

In fact, since joining the Photosynthetic Organization, Trend Technology has carried out a series of ecological cooperation with upstream and downstream partners in the industry chain such as Haiguang. In Trend Technology's view, what users need is a complete solution, and it is difficult for a single manufacturer to provide a complete set of products from the underlying architecture to front-end applications. In this context, the integration of ecological forces is very critical.

"Photosynthetic Organization provides a good platform for ecological docking of the AI ​​industry." Wen Yuxiang pointed out that based on the Photosynthetic Organization Competition, Trend Technology not only deepened product technology cooperation with partners and demonstrated program results, but also Extensive linkage with more industrial forces has gradually formed an ecological "flywheel", and has gradually accelerated in areas such as market empowerment and project introduction.

"I believe that as the speed of the flywheel continues to accelerate, in the future, the computing power pooling solution will reach more business application scenarios, and AI computing power will release greater value." Wen Yuxiang said.

The above is the detailed content of Trend Technology: Resource pooling releases the value of AI computing power. For more information, please follow other related articles on the PHP Chinese website!

source:sohu.com
Statement of this Website
The content of this article is voluntarily contributed by netizens, and the copyright belongs to the original author. This site does not assume corresponding legal responsibility. If you find any content suspected of plagiarism or infringement, please contact admin@php.cn
Popular Tutorials
More>
Latest Downloads
More>
Web Effects
Website Source Code
Website Materials
Front End Template