Many supercomputer centers are fast-tracking COVID-19-related proposals and prioritizing COVID-19-related jobs on their systems. For the researchers whose access to these systems is limited (or for whom time is too limited), cloud-based HPC systems offer a convenient solution. Chinese commerce and cloud giant Alibaba recently highlighted how researchers fighting against the coronavirus have leveraged its cloud services to enable necessary HPC and AI infrastructure for their research.
In mid-March, Alibaba Cloud, like many cloud service and software companies, offered COVID-19 researchers free access to a number of its tools. Chief among them was access to Alibaba Cloud’s Elastic High-Performance Computing (E-HPC) Solution for Life Sciences. Alibaba’s E-HPC offering, which offers flexible cluster management and auto-scaling, includes four Intel CPU-based configurations (mostly based on the Xeon E5-2682) and one similar configuration with additional Tesla P100 GPUs, all with network speeds ranging from 10 to 25 Gbps.
In a presentation, Warren He Wanqing, director of cloud HPC, elaborated on how cloud HPC tools have been leveraged by COVID-19 researchers. According to He Wanqing, the work can be traced back to mid-January, when Alibaba began working with the Global Health Drug Discovery Institute (GHDDI) in Beijing. As the pandemic spread, Alibaba saw researchers using its services more and more, expanding quickly to 14 public research agencies using Alibaba Cloud – and now, around 20.
Most of the organizations were universities in the region (including Zhejiang University, Shenzhen University, Macao University of Science and Technology and Tsinghua University), but others – like GHDDI and XtalPi, a pharmaceutical technology company – have leveraged the services as well.
He Wanqing highlighted two main uses of Alibaba’s E-HPC offerings for COVID-19: drug discovery and diagnosis. By way of example, he focused on GHDDI, which used cloud HPC to create an AI-driven drug discovery platform. Using Alibaba-provided systems, GHDDI made its computational work on COVID-19 available to fellow researchers on the cloud, including a massive data repository of antiviral pharmacology results, 3D structure data for COVID-19, virtual libraries of compounds numbering in the millions, trained AI-based drug screening models, a PyTorch-based deep learning environment, parallel processing-enabled AutoDock and more.
Then, He Wanqing discussed diagnosis, which (in China) has largely been conducted by analysis of CT scans. Unfortunately, each CT diagnosis takes a considerable amount of time for a doctor in a time when doctors are in short supply. To bridge this gap, HPC-enabled AI has stepped in, analyzing in minutes – or even seconds – what it might take a doctor 15 minutes to assess. Using Alibaba’s E-HPC, Tsinghua University implemented one of these solutions, completing diagnoses in just ten seconds. An ensuing Alibaba product, CT Image Analytics, was made free to research institutions and is now deployed at over 160 hospitals in China.
He Wanqing mentioned other use cases, as well, including Sun Yat-Sen University (where E-HPC was used to accelerate data transfers fivefold to enable research on COVID-19’s evolution) and Peking University (where E-HPC was used to accelerate molecular docking tests). The list goes on – as can be seen below – and Alibaba appears committed to ensuring that cloud HPC is a major player in the COVID-19 era.