Fujitsu introduces dynamic resource allocator for AI servers and HPC systems

Forward-looking: Fujitsu’s business has traditionally focused on selling computing products, mainframe servers, and HPC devices. Now, the company is aiming to leverage its “computational optimization expertise” to offer innovative, software-based solutions to address the growing demand for GPUs in cutting-edge AI applications.

After introducing its “computing broker” solution in 2023, Fujitsu has now confirmed that the product is finally available for purchase in Japan and other markets worldwide. The Kawasaki-based corporation aims to achieve through software what has traditionally been handled by hardware, which is expected to significantly improve resource optimization and GPU utilization.

The new technology is presented as middleware designed to dynamically allocate resources on a per-GPU basis, optimizing utilization and advanced memory management across multiple platforms and AI applications. The computing broker allocates CPU and GPU computing resources in real-time, prioritizing processes with higher execution efficiency. It can also reallocate processes even while they are already running on a GPU.

In pre-release testing, Fujitsu reportedly achieved up to a 2.25x improvement in GPU processing performance. The technology also offers impressive memory management capabilities, as it is designed to handle AI workloads of up to 150GB – around five times the physical memory capacity of the tested GPUs.

Fujitsu initially stated that developers needed to use its proprietary framework to fully exploit the capabilities of the new computing broker technology. However, the latest announcement makes no mention of this requirement. The company is now working to further enhance the technology, aiming to support multiple GPUs installed across multiple servers in data center environments.

Fujitsu said that various companies have been testing the computing broker middleware since May 2024. Tradom, a Japanese fintech venture, has reportedly implemented the technology in production, while cloud provider Sakura is evaluating its potential for optimizing data center operations.

Fujitsu emphasizes the value of resource optimization in reducing the energy consumption of GPU-based AI applications. With generative AI services continuing to dominate the tech landscape, enterprise-grade GPUs remain among the most in-demand hardware components. The company suggests that making these systems operate more efficiently is crucial for addressing the growing demand.