Hewlett Packard Enterprise has announced the construction of a new supercomputer for Singapore’s meteorological agency, Meteorological Service Singapore (MSS, itself a division of Singapore’s National Environment Agency). The new system affords MSS almost twice the computing power relative to its Cray-based predecessor, which MSS says will enable important forecasting and modeling improvements.
The system – which does not have a name – is based on the HPE Cray supercomputer and powered by 196 AMD Epyc “Milan” CPUs. Networked with HPE Slingshot 10, the system is complemented by a Cray ClusterStor E1000 parallel storage system. The system, hosted at MSS’ Centre for Climate Research Singapore (CCRS) datacenter, is already operational and has already been accepted. It delivers 401.4 peak teraflops.
While those teraflops might be shy of the Top500 list’s entry point, that supercomputing power will enable important capabilities for MSS. The system will accelerate SINGV, MSS’ numerical weather prediction system (developed in collaboration with the UK’s Met Office); it will enable machine learning-based forecast postprocessing algorithms; and it will power the development of a sub-kilometer urban weather and climate model, as well as a coupled ocean-atmosphere-land-wave modeling system that MSS is calling cSINGV – and which, of course, tackles interactions of particular relevance to Southeast Asia.
“At CCRS, our scientists and software engineers are committed to developing advanced modeling systems and examining complex data to provide timely weather forecasts for our nation, which due to the island’s unique geological positioning, often experiences various weather processes on a daily basis,” said Dale Barker, director of CCRS. “After collaborating with Hewlett Packard Enterprise to design the new supercomputer, our research center will gain a faster system with next-generation technologies to advance modeling and simulation tools, while introducing new capabilities to test and apply future types of applications for deeper research methodologies.”
The new computational resource marks another major supercomputing investment for Singapore, which has worked hard to establish an outsized HPC presence for a relatively small country. HPE also built a much larger system for Singapore’s National Computing Center (NSCC), also powered by Milan CPUs (as well as Nvidia A100 GPUs). That system’s GPU partition weighs in at 2.86 Linpack petaflops, while the CPU partition weighs in at 2.58 Linpack petaflops. Respectively, they rank 199th and 247th on the Top500. The aggregate peak of the partitions is 8.23 petaflops.
“Singapore continues to drive its national initiatives with supercomputing projects that significantly contribute to science, accelerate innovation and improve a range of areas for the greater good of its citizens,” said Trish Damkroger, chief product officer and senior vice president for HPC, AI, and Labs at HPE. “We are honored to play a role in the nation’s digital agenda and be selected by the Meteorological Service Singapore to build them a powerful system with advanced, end-to-end supercomputing technologies that will fuel Singapore’s weather intelligence and speed up predictions of extreme climate events.”
NSCC also recently partnered with Nvidia and SingHealth to make “supercomputers and software” available to power deep learning and AI applications in Singapore’s healthcare and medical research initiatives. More on that here.