TOKYO, Oct 22, 2024 – (JCN Newswire) – Fujitsu today announced the launch of an AI computing broker middleware technology designed to optimize GPU performance for AI processing and address the global GPU shortage. The new technology combines Fujitsu’s proprietary GPU allocator technology, which allocates GPUs for real-time use, with a number of AI optimization techniques.
After successful pilot tests, TRADOM Inc. (1) will start using AI computing broker technology solutions in October 2024. In addition, SAKURA internet Inc. (2) has begun to explore the possibilities of AI computing broker technology for its data center. work.
Since May 2024, Fujitsu has been testing this new technology with AWL, Inc., (3)Xtreme-D Inc., (4) and Morgenrot Inc., (5) and has shown significant improvements in these industries. ‘work. These tests confirmed a 2.25x increase in the use of different AI methods and a significant increase in the number of AI methods used simultaneously in different cloud environments and servers.
The newly developed technology will be available to Japanese customers from October 22, 2024 and to users worldwide.
Fujitsu will continue to offer its AI computing broker technology to end users including AI service providers who want to reduce GPU costs by improving performance, and to cloud service providers who want to expand GPU usage. Addressing the challenges of GPU shortages and power consumption driven by the global demand for AI, Fujitsu aims to help improve business and technology services for its customers.
Dealing with the growing power of AI and GPU distribution
Driven by the increasing global demand for AI technology (including AI output), the demand for GPUs, which are more suitable for AI processing than CPUs, has grown significantly. The global market for generative AI is expected to grow approximately 20% from 2023 to 2030 (6), and as such, the demand for GPUs is expected to rise accordingly. However, the increase in power consumption in data centers to meet the demand for GPUs poses a major challenge. It is estimated that data centers will consume 10% of the world’s electricity by 2030 (7).
To solve this global problem, Fujitsu developed the GPU allocator technology in November 2023. The technology maximizes the use of CPUs and GPUs by allocating resources in real-time to give priority to efficient processes, even when the GPU is active. program. Fujitsu has been conducting validation tests of this distribution technology on various platforms.
About the new founder of AI computing broker
The newly developed AI computing broker middleware combines GPU processing technology with AI optimization technologies, predicting and optimizing GPU resource allocation for AI processing across multiple applications.
Unlike conventional distribution for each application, Fujitsu’s AI computing broker allocates GPU resources for each GPU model, leveraging Fujitsu’s expertise in technology optimization to improve availability. The ability of GPU technology allows users to run many AI processes without worrying about using GPU memory or physical power.
In pre-release tests, Fujitsu’s AI computing broker showed a 2.25x improvement in GPU performance in terms of time per session compared to rendering without the technology. In addition, the advanced memory management enables simultaneous operation of up to 150GB of AI processing, approximately five times the amount of GPU memory.
Going forward, Fujitsu will expand the scope of its AI technology project, including installing multiple GPUs installed on multiple servers, in anticipation of being used in large computing environments. Fujitsu will continue to develop advanced computing technologies to overcome challenges including GPU and power shortage, contributing to the realization of AI that increases productivity and technology as a reliable assistant.
Junichi Kayamoto, Chief Data Science Officer, TRADOM Inc.
“Our company provides advanced AI solutions for foreign currency risk management. Fujitsu’s AI computing broker’s technology testing proved its ability to optimize the GPU parameters of AI models, which helps to create more accurate models in less time through AI learning process. We are committed using this technology, by continuing the collaboration with Fujitsu, to increase our solutions, and drive the growth of TRADOM and the progress of the FinTech industry.”
Ken Wakishita, Senior Director, SAKURA Internet Inc. / SAKURA Internet Research Center comments:
“Our testing of Fujitsu’s AI computing broker has proven the technology’s ability to improve the performance of GPUs in our cloud business, expanding GPU access to more customers. We hope to work with Fujitsu to integrate this technology and achieve more GPUs.”
Hiroshi Fujimura, R&D General Manager – AWL, Inc., comments:
“We are dedicated to developing and providing AI camera solutions that solve various problems and increase the demand of real world environments, especially commercial environments. To meet the great demands of our customers, optimizing the cost of the GPU during the training of the AI ​​model is essential. Our tests confirmed the potential of Fujitsu’s AI computing broker to advance the use of GPU and AI processing. We are excited to see the progress of this important technology.”
Naoki Shibata, Representative Director, CEO and CTO, Xtreme-D Inc., comments:
Congratulations to Fujitsu for the launch of AI computing broker. Xtreme-D offers Raplase, a cloud-based service for AI and HPC customers. A major challenge for our customers is cost/performance optimization by effectively using expensive and bare metal cloud GPUs. We are confident that Fujitsu’s AI computing solution will solve this problem, and we are working closely with Fujitsu to integrate this solution into our customer offerings. “
Masamichi Nakamura, COO, and Hisashi Ito, CTO, Morgenrot Inc., comment:
“Our company is changing the landscape of cloud computing with its advanced and innovative solutions for data storage. We are creating an economic solution for sharing computing power, ready to meet the needs of modern products, including GPUs. Our latest experiment, with the aim of combining HPC management solution (M : Arthur) and the cloud service (Cloud Bouquet) with Fujitsu’s AI computing broker technology, gave impressive results. By allowing the sharing of the GPU between several tasks, we achieved a remarkable reduction of about 10% in the total processing time compared to running sequentially on the machines. This parallel capability opens up many advantages, allowing for the simultaneous execution of long prototyping and short simulation/testing tasks, all within a limited amount of resources. We are excited to explore how Fujitsu’s AI computing broker is used to integrate this revolutionary technology into our products. “
[1] The opinion of the company TRADOM Inc.Headquarters: Otemachi, Chiyoda-ku, Tokyo, Japan; President and co-CEO & CTO: Shinichiro Urashima and co-CEO & CSMO Shin Sakanehttps://www.tradom.jp/company-en
[2] SAKURA internet Inc.’s opinionHeadquarters: Kita-ku, Osaka-shi, Osaka, Japan; Founder & CEO, President: Kunihiro Tanakahttps://www.sakura.ad.jp/corporate/en/
[3] Company opinion AWL, Inc.Headquarters: Chiyoda-ku, Tokyo, Japan; President and CEO: Muneharu Kitadehttps://awl.co.jp/en/
[4] Opinions of the company Xtreme-D Inc.Headquarters: Shinagawa-ku, Tokyo, Japan; Founder and CEO: Naoki Shibatahttps://xtreme-d.net/ (in Japanese)
[5] The opinion of the company Morgenrot Inc.Headquarters: Chiyoda-ku, Tokyo, Japan; Representative Director, CEO: Ryuei Morimotohttps://morgenrot.net/
[6] The global generative AI market is expected to grow at a CAGR of 20% from 2023 to 2030:Source: Japan Electronics and Information Technology Industries Association “JEITA Announces Global Demand Forecast for Artificial Intelligence” (December 21, 2023) (in Japanese)
[7] It is estimated that data centers will use 10% of the world’s electricity by 2030:Show February 2020 Japan Science and Technology Agency “Impact of Information Society on Energy Consumption (Vol.2)” (in Japanese)
About Fujitsu
Fujitsu’s mission is to make the world sustainable by creating trust among people through innovation. As a partner in the digital transformation of customers in more than 100 countries, our 124,000 employees work to solve the biggest problems facing society. Our diverse services and solutions are based on five key technologies: Computing, Networks, AI, Data & Security, and Converging Technologies, which we bring together to bring about lasting change. Fujitsu Limited (TSE:6702) reported consolidated revenue of 3.7 trillion yen (US$26 billion) for the fiscal year ended March 31, 2024 and remains Japan’s top digital technology company by market share. Find out more: www.fujitsu.com.
Tap Contacts
Opinions of the company Fujitsu Limited
Public and Investor Relations Division
Questions
#Fujitsu #introduces #computing #broker #middleware #address #global #GPU #shortage #improve #performance