“Advances in IT are making HPC systems increasingly more powerful and innovative to accelerate the time necessary to reach new discoveries, but many still believe implementations can be complex,” said Thierry Pellegrino, vice president and general manager of HPC at Dell EMC. “Based on decades of experience with leading institutions, technology partners and strategic customers, Dell EMC provides an extensive portfolio of technologies that simplify HPC adoption to advance research and further democratize HPC. We remain focused on leading the way in HPC innovation and helping organizations of all types and sizes further advance expanding opportunities in artificial intelligence and machine learning.”
Dell EMC fueling research for human progress
Dell EMC continues to be at the forefront of helping customers adopt the latest HPC technologies to fuel a wide range of discoveries and research. Recent customer momentum demonstrates Dell EMC’s commitment to deliver world-class HPC systems that bring together the latest advances in servers, accelerators, liquid cooling and networking:
- Texas Advanced Computing Center (TACC) at the University of Texas at Austin has selected Dell EMC to develop and deliver its new Frontera supercomputer in 2019, funded by TACC’s $60 million award from The National Science Foundation. At the time of its announcement, in August 2018, Frontera would have been the world’s fifth most powerful system, the third fastest in the U.S. and the largest at any university if completed. The Dell EMC PowerEdge system plans to combine several technical innovations such as CoolIT Systems high-density Direct Contact Liquid Cooling, high performance Mellanox HDR 200Gb/s InfiniBand interconnect and next generation Intel® Xeon® Scalable processors. Frontera’s early projects are expecting to include analysis of particle collisions from CERN’s Large Hadron Collider, global climate modeling, hurricane forecasting and multi-messenger astronomy.
- The University of Cambridge has expanded its supercomputing capabilities with its “Cumulus – UK Science Cloud.” This new OpenStack system is the UK’s largest academic supercomputer, providing more than two petaflops of performance, powered by Dell EMC PowerEdge servers, Intel® Xeon® processors and Intel® Omni-Path Architecture. To help solve the UK’s most challenging data driven, simulation and AI tasks, Cumulus is open to all UK academics and industry and delivered in partnership with Dell EMC and StackHPC, a UK start-up specialising in the convergence of HPC and Cloud. It is funded with investments totalling over £13 million from STFC (DiRAC/IRIS), EPSRC (Tier 2) and the university.
- The University of Michigan is deploying its Great Lakes computing cluster for simulation, modeling, artificial intelligence, machine learning, data science, genomics and more. The new system is powered by a Dell EMC-enabled HPC infrastructure built on Dell EMC PowerEdge servers. Great Lakes is the industry’s first system to benefit from Mellanox HDR 200Gb/s InfiniBand networking, enabling faster data transfer speeds and increased application performance.
- The Ohio Supercomputer Center is deploying its Pitzer Cluster, delivered by Dell EMC. Like TACC’s Frontera system, the Pitzer Cluster will utilize Dell EMC PowerEdge servers with CoolIT’s modular, rack-based Direct Contact Liquid Cooling solution, which allows for increased rack densities, higher component performance potential and better energy efficiency. As a result, it will offer nearly as much performance as the center’s most powerful cluster but require less power and less than half the space. The system will power broad research areas from human genomics to the global spread of viruses.
Dell EMC eases HPC adoption with Ready Solutions advancements
Today’s HPC workloads require storage infrastructure that scales endlessly and delivers unmatched bandwidth at high concurrency for deep learning algorithms and AI initiatives. To meet these needs, Dell EMC is committed to expanding its HPC portfolio to offer a range of high performance storage options that complement its portfolio of Ready Solutions with the Dell EMC Isilon Scale-out NAS storage powered by the Isilon OneFS operating system.
The Dell EMC Ready Solution for HPC Lustre Storage and Dell EMC Ready Solution for HPC NFS Storage are now available with the new Dell EMC PowerVault ME4 storage arrays. Dell EMC built the ME4 Series with 75 percent more drives than the PowerVault MD3 to increase raw storage capacity by 122 percent, while also boosting read IOPS performance by 4X. Its modular design allows for flexible and custom designs, offering increased density, when compared to the PowerVault MD3, and the ability to scale as customers’ businesses grow.
Ideal for technical, big data applications, the Dell EMC Ready Solution for HPC Lustre Storage with the new PowerVault ME4 delivers excellent throughput per building block with on-the-fly storage expansion. As an easy-to-use and fully redundant NFS storage solution, optimized for HPC environments, the Dell EMC Ready Solution for HPC NFS Storage on ME4 Series will offer greater overall performance and a denser solution. Both solutions are accompanied by Dell EMC global services and support.
Dell EMC PowerEdge servers to support latest accelerator technology
Dell EMC PowerEdge R640, R740, R740xd and R7425 servers will support the latest GPU and FPGA accelerators to speed results in data analytics and AI applications. This includes:
- NVIDIA® Tesla® T4, the universal AI accelerator ideal for distributed computing environments. It is packaged in an energy-efficient 70-watt, small PCIe form factor. Powering breakthrough performance, NVIDIA notes that Tesla T4 provides multiple times the performance of traditional CPUs for both training and inference.Developers can unleash the power of NVIDIA Turing architecture-based Tensor Cores directly through NVIDIA TensorRT and cuDNN software libraries and integrations with all AI frameworks.
- From video content streaming to financial services to defense applications, FPGAs allow hardware to be programmed and re-programmed for optimization. In addition to the Intel® Arria® 10 GX FPGA support today, Dell EMC is now the first server vendor to qualify the Xilinx® Alveo™ U200 accelerator card, adding it to Dell EMC PowerEdge accelerator options. Xilinx notes, for machine learning, Alveo accelerators can increase real-time inference throughput for machine learning by 20X versus high-end CPUs alone.
Dell EMC AI Challenge winner
Dell EMC has selected a research team at the Center of Space, High-Performance, and Resilient Computing (SHREC) at the University of Florida as the winner of the 2018 Dell EMC AI Challenge. The AI Challenge, launched in May 2018, encouraged entrants to demonstrate practical applications of AI technology with a transformational impact on business, research or society. The winner receives 200,000 core-hours on the Dell EMC HPC & AI Innovation Lab Zenith cluster and a spotlight in Dell EMC’s booth at the SC18 conference among other promotional activities
SHREC is comprised of more than 30 industry, government and academic partners working together to solve research challenges in missions and applications that drive and can benefit from reconfigurable, high-performance and reliable computing. For the AI Challenge, the SHREC team was recognized for developing and demonstrating a heterogeneous computing (HGC) system that can support a complete workflow–data analysis and pre-processing, model training, deployment and inferencing–for machine learning and be applied to any application domain leveraging machine learning, including healthcare, business, finance, science exploration and more.
“For the AI Challenge, our team leveraged CERN OpenLab datasets to determine the performance of the HGC workflow with CPUs, GPUs and FPGAs for machine learning. The study showed performance gains of 1.45-2.22x,” said Chao Jiang, Ph.D. student leader of the team. “These early results were promising, and we are continuing to experiment with more complex 3D-image based techniques such as volumetric segmentation with 3D U-net to improve performance, as well as 3D GAN for accelerated particle-simulation.”
For more information, please visit www.dellemc.com/hpc
