Fujitsu is pushing the boundaries of high performance computing (HPC) with a newly-developed hardware cluster that will increase the pace of scientific research and lead to better and faster understanding of complex systems, such as weather forecasts.
The platform’s computing performance has been successfully validated at EPCC, the University of Edinburgh’s supercomputing center, and handed over by Fujitsu for release to operations.
Fujitsu created the prototype in a close co-creation process with hardware and software vendors, universities and research institutes as part of the NEXTGenIO research and development project funded by the European Commission’s Horizon 2020 program.
The objective of the NEXTGenIO project is to solve this input/output (I/O) bottleneck by designing and developing a new, scalable, high-performance and energy-efficient computing platform capable of delivering scalable I/O performance to applications at the Exascale.
Following testing and validation, a prototype 34-node computing cluster, designed by Fujitsu using Intel components, has now been handed over and released to operations at EPCC’s Advanced Computing Facility at the University of Edinburgh.
Focusing on maximum performance, Fujitsu developed the NEXTGenIO system and systemware to close the gap between memory and storage by leveraging Intel Optane DC Persistent Memory Modules. Each of the 34 compute nodes is equipped with two second-generation Intel Xeon Scalable Family processor CPUs and 3TB of Intel Optane DC persistent memory, and includes a software stack to seamlessly support I/O and memory intensive workloads.
Ingolf Stärk, senior director: HPC at Fujitsu, says: “We are delighted to deliver the results of a very successful co-design process. Together with the other project partners, we have created an ultra-high-performance computing platform that will pave the way for the extremely data-intensive, next-generation computing challenges of the future. Fujitsu is committed to driving the joint development of the European ecosystem for high performance computing.”
Professor Mark Parsons, EPCC direct and project co-ordinator of NEXTGenIO, says: “We are proud to be hosting and operating the NEXTGenIO prototype cluster, which will allow us to push the boundaries of research into memory and I/O bottlenecks often experienced by HPC and high performance data analytics applications. The collaboration with Fujitsu in particular has been excellent, and it has been exciting to witness the progress all the way from leading the design of the architecture to taking delivery of the hardware, which we are sure will be transformational for many applications.”
The NEXTGenIO consortium partners have already demonstrated impressive performance improvements with the new computing technology.
* ECMWF, the European Centre for Medium-Range Weather Forecasts, is one of the end users represented in the NEXTGenIO project. ECMWF prepares its weather forecasting systems for the massively parallel supercomputing facilities of the future and is reporting the NEXTGenIO prototype shows an increase of end-to-end bandwidth with the distributed field data base of about factor 10, in comparison with today’s conventional I/O system based on Lustre/HDD.
* The efficiency of the utilization of compute resources and the scientific throughput is increased with the NEXTGenIO system when tested with CASTEP – a leading simulation code for materials science – opening up the potential for rapidly accelerated research projects across multiple domains. For a DNA simulation with CASTEP, a relative efficiency improvement of about 2x has been achieved
Rupert Lehner, head of Central and Eastern Europe: EMEIA Products, says: “In close cooperation with our project partners, we have achieved a true breakthrough in non-volatile memory technology. This research will enable us to build an entirely new generation of computing platforms with the speed and performance required to power tomorrow’s data-intensive and real-time applications. After testing and optimizing the platform in a range of high-performance scientific computing applications, we are looking forward to passing on these significant computing advances to our customers and partners.”
The NEXTGenIO prototype represents the culmination of more than three years of collaborative research between hardware and software vendors, universities and research institutes. Fujitsu has now made technology developed and validated within the project available to its customers in its Primergy and PrimeQuest servers.