Purdue computer scientist searching for ‘buried treasure’ with AMD EPYC Processor-based Microsoft Azure hybrid cloud extension to community clusters
A Purdue professor who searches for factors of very large numbers – work that is essential to keeping everything from nuclear weapons to personal banking information secure - is using new hybrid cloud capabilities between Purdue community HPC clusters and the Microsoft Azure HPC cloud in his hunt.
Samuel Wagstaff, a professor of computer science, uses computers to find factors for 200-digit numbers and to determine whether large numbers are prime, that is, whether they have no factors other than 1 and the number itself.
The factorization of large numbers has important applications to the field of cryptography, because prime numbers are used to create cryptographic keys that are widely used in information security. Choosing the right size prime is a balancing act – too small and the code can be broken easily; too large and communication is unnecessarily slowed.
Wagstaff likens the search for factors of large numbers to a search for treasure buried deep beneath a field. The longer a computer can run a job, the “deeper” a computer scientist can dig for treasure.
Most of the “treasure” buried just beneath the surface has already been found, meaning scientists like Wagstaff have to dig deep to find anything new. That’s where Azure HPC comes in.
Using HBv3 instances on Azure featuring 3rd Gen EPYC CPUs, Wagstaff is able to dig three to four times deeper than he could using standby queues on the community clusters, meaning that he is 10 times as likely to find a factor on each search. Azure also offers access to very large capacities and scales for higher throughput searches and reduced time to results.
The Azure HPC environment has been set up by Research Computing staff to extend the Purdue community HPC cluster environment, meaning users who are familiar with the community clusters won’t have to make major adjustments to existing workflows.
“Researchers can move from an on-premises system to a cloud system without having to make any modifications to their code,” explains Erik Gough, lead computational scientist for Research Computing, who has led the Azure project for Purdue Research Computing. “This will allow Purdue researchers easy bursting to meet deadlines, or access to specialized instances not available at Purdue.”
The fact that Azure has been integrated into Research Computing’s existing infrastructure means that Research Computing is able to support researchers awarded grant funding or credits for a public cloud computing resource. Supported by the Microsoft Azure HPC & AI Collaboration Center this new capability will also allow users of the NSF-funded Anvil system to burst to AMD EPYC CPU based HBv3 instances in the Azure HPC cloud.
To learn more about how researchers can use Azure or other Research Computing resources, contact firstname.lastname@example.org.
Writer: Adrienne Miller, science and technology writer, Research Computing, email@example.com.