The institute's goal is to make precision medicine — the right treatment for the right patient at the right time — real, explained Garvan Institute head of infomatics Warren Kaplan.
This requires the study of disease at scale which, in turn, requires large amounts of computational power and data storage.
"The entire demographic has changed," said Kaplan, who joined the institute as its first and only bioinformatician. Now, 80 are on the payroll and the role has changed from being seen as a service provider to becoming an active part of research teams.
|
|
Garvan's new Dell EMC HPC infrastructure features 14th generation PowerEdge servers with Nvidia GPUs and Intel Xeon processors and FPGAs (field-programmable gate arrays).
Specifications include:
• 47 Dell EMC PowerEdge servers with 1,632 Intel Xeon Scalable Processor cores
• 10 Intel Arria 10 GX FPGAs
• 122,880 Nvidia Tesla V100 CUDA cores
• 15,360 Nvidia Tensor cores
• 744TB of NVMe
• 41TB DDR4 RAM
• 530TB usable capacity CephFS storage
• 25GbE and 100GbE interconnects
The new equipment will support traditional computational analysis and simulation while adding extensive big data analytics and deep learning capabilities.
The FPGAs can be used to accelerate various functions. In particular, some parts of version 4 of the Genome Analytics Toolkit (GATK; one of the pieces of software used by the institute) includes FPGA support, said Kaplan.
Dell EMC ANZ high performance computing lead Andrew Underwood said most of GATK would take advantage of FPGAs by the end of 2018, and consequently a large proportion of genome sequencing could be done on FPGAs in future.
The new Garvan system is "the first of its kind in Australia," and is thought to be the first in the APAC region (Dell EMC is one of only two OEMs using Intel's FPGAs), he said.
Furthermore, Dell and Intel will connect Garvan Institute with relevant software developers in "a true partnership," he added.
In any case, Underwood said staff have other ideas about how FPGAs can be used in their work.
"There's so much opportunity" to explore possible applications of IT to medical research, even though not all of it will get as far as production," he said.
"We're as enthusiastic as everybody" when it comes to investigating use cases. For example, the upgrade includes a small number of GPU nodes, and "we'll see how this goes". Early enthusiasm for GPUs faded, but he expects the technology to become more pervasive.
The new additions to Garvan's infrastructure took about two months to install alongside the existing equipment (which remains in use, despite being up to six years old), largely because of the extra power and cooling required. But from a systems point of view, adding the new nodes was straightforward, said Kaplan.
Garvan's genome sequencing capability is used by the Kinghorn Centre for Clinical Genomics, the Garvan Weizmann Centre for Cellular Genomics, the Sydney Genomics Collaborative (including the Medical Genome Reference Bank, which contains 4000 whole genome sequences to use as controls in disease research), the Lions Kids Cancer Genome Project, and the Australian Genomic Cancer Medicine Program (which seeks to uncover new treatments for rare cancers).
"We want to change the direction of medicine and have a life-changing impact on people's health. We see genomics as the key to driving this transformative change, and we couldn't achieve this without the computing infrastructure to make it possible," said Kaplan.
Disclosure:The writer attended the Dell Technology Forum as a guest of the company.
