Microsoft Azure deployed a cloud supercomputer with a performance of almost 5 Pflops / ServerNews

Microsoft Azure deployed a cloud supercomputer with a performance of almost 5 Pflops / ServerNews


A new high performance computing (HPC) scaling record has been achieved in the Microsoft Azure cloud. There we managed to create a virtual cluster with 86 400 CPU cores. For this, physical servers based on AMD EPYC 7002 with HDR InfiniBand connection were used, which were combined into 720 Azure HBv2 virtual machines. This resulted in a peak performance of 5 Pflops.

The cloud provider partnered with the Beckman Institute of Advanced Science and Technology at the University of Illinois at Urbana-Champaign (UIUC) to test the limits of Azure’s scalability. The workload was the molecular dynamics modeling tool (NAMD) developed at UIUC. The tool was instrumental in this year’s COVID-19 research.

The team found that not only did the HBv2 clusters meet the needs of the researchers, but that the performance and scalability in Azure rivaled and in some cases surpassed the capabilities of the # 8 Frontera supercomputer on the June 2020 Top500 list.“Said Evan Burness, Chief Specialist and Program Manager for Azure HPC.

This will allow the power of the cloud to be used for future projects. In simple terms, this could allow faster cure for coronavirus and other diseases to be found. And David Hardy, Senior Research Programmer at UIUC, noted that the Charm ++ library, which allows parallelization of tasks, has achieved good performance and scalability for NAMD on a Microsoft Azure cluster. Moreover, this did not require rewriting the source code.

Recall that earlier in Azure they showed a new record for I / O performance on the BeeGFS file system.