Intel has spun out the Omni-Path business to Cornelis Networks, an independent Intel Capital portfolio company. Cornelis Networks will continue to serve and sell to existing and new customers by delivering leading purpose-built high-performance network products for high performance computing and artificial intelligence. Intel believes Cornelis Networks will expand the ecosystem of high-performance fabric solutions, offering options to customers building clusters for HPC and AI based on Intel® Xeon® processors. Additional details on the divestiture and transition of Omni-Path products can be found at www.cornelisnetworks.com.
Executive Summary
As a leading education and research institution in China, Peking University embraces many branches of learning in applied and pure sciences, social sciences, humanities, health and medicine, business, and education. Important to its research support is the university’s high performance computing (HPC) facility. Peking University’s HPC resources help scientists make new discoveries in materials, physics, chemistry, weather and climate, and many other disciplines. Its latest additions to the HPC facility include two clusters for teaching and student research based on Intel® Xeon® Scalable processors and Cornelis Networks fabric.
Challenge
HPC facilities are an important component to academic research. To maintain its reputation as a leading research institution in China, Peking University needed greater computing capability in its HPC facilities to support ongoing research in chemistry, image processing, artificial intelligence (AI), cryo-electron microscopy (cryo-EM), and other sciences.
Solution
In 2018, Peking University turned to supercomputing vendors for new supercomputing clusters. In May, the university put into production a four-node teaching cluster for education. This system was built with Inspur NF5280M5 servers using Intel® Xeon® Gold 6132 processors with 14 cores each. The total cluster provides 112 cores, which are used to create many virtual machines across the system.
Later that year in October, a much larger research cluster called Weiming Life Science #1 was put into production with 150 total nodes. 129 compute nodes of Lenovo SD530 servers each house two Intel® Xeon® Gold 6142 processors with 16 cores, and seven Lenovo XS1500 GPU nodes each include two Intel® Xeon® E5-2690 v4 processors and four graphics cards. An additional 14 nodes of Lenovo SR650, SR630, and Lenovo System X3650 M5 servers provide login, management, and storage services. The compute and GPU nodes are interconnected by 100 Gbps Intel Omni-Path Architecture (Intel OPA).
The research cluster delivers a total of over one-half petaFLOPS of peak performance (587.8 teraFLOPS).
Peking University’s HPC resources help scientists make new discoveries in materials, physics, chemistry, weather and climate, and more.
Results
“The new clusters provide the high-performance scientific and engineering computing services needed across the entire university,” commented Mr. Chun Fan, Senior Engineer, Chief of the System Management Section in Computing Center Peking University and Chief Engineer in HPC platform of Peking University. “These systems meet the needs of large-scale data processing and scientific computing required by research in cryo-EM, deep learning, biomedical, physics, and other sciences.”
Many projects have already taken advantage of the computing capabilities of Weiming Life Science #1, such as the following:
- Computational support and visualization of cryo-EM research on Eukaryotic cell structure and interaction with human proteasome. (Cryo-EM Structures and Dynamics of Substrate-Engaged Human 26S Proteasome)
- Creation of an objective method of Surveillance Video Quality Assessment (SVQA) of images from surveillance data. (How to Assess the Quality of Compressed Surveillance Videos using Face Recognition)
- STEM and Ab initio simulation of methylammonium lead iodide perovskite research for next-generation solar cell development. (Atomic scale insights into structure instability and decomposition pathway of methylammonium lead iodide perovskite)
These and many other reports are available on the Peking University HPC’s web site.
Solution Summary
- Intel Xeon Gold and Intel® Xeon® Silver processors
- 150-node Weiming Life Science Cluster #1 built by Lenovo
- Intel Omni-Path Architecture fabric
- 587.8 teraFLOPS peak performance