The University of British Columbia Drives Scientific Research Without Limits Using RONIN on AWS

The University of British Columbia Advanced Research Computing leverages the RONIN HPC platform to access ondemand compute resources, quickly spin up an 8,640-core cluster, and drive new scientific research. RONIN uses Intel powered Amazon EC2 instances and takes advantage of AWS ParallelCluster to automatically launch HPC environments.

One of Canada’s leading public research institutions, The University of British Columbia (UBC), conducts innovative research in science, medicine, social sciences, and the arts. UBC researchers attract funding from government, industry and non-profit sources, with the annual research budget of more than CAD $750 million supporting over 10,000 projects. UBC Advanced Research Computing (ARC) provides high performance computing (HPC) infrastructure and computational support to researchers across all disciplines. “We support impactful research, which has included the development of AI and machine learning for self-driving cars, and the use of large volumes of social media data to try and map the spread of COVID-19,” says Matthew Smith, manager of research systems with the University of British Columbia’s Advanced Research Computing unit.

UBC has invested in on-premises HPC and storage resources for its research community. However, as the university’s research computing needs continue to grow, the finite nature of the available resources can be a challenge. For instance, researchers could often wait several days or longer in queues to use HPC resources, which all researchers share. Scalability is another limitation. “Some of our on-premises compute resources are finite enough that the HPC environment doesn’t allow people to accomplish their research within the desired time frame,” Smith says. “Sometimes there are not enough computational cycles for them to meet conference or publication deadlines. We were interested in finding a cloud-based HPC environment to solve these types of critical problems.”

The ARC team reached out to RONIN, an AWS HPC Competency Partner, after seeing a demo of the AWS-based RONIN self-service HPC platform. RONIN is a web application that allows researchers to take advantage of AWS infrastructure without having to learn the complex details of cloud computing. “We were really interested in the ease of use of RONIN,” Smith says. “It is important for us that even researchers with very little experience should be able to access a tool such as RONIN and to create their own HPC cluster very quickly. As administrators, RONIN has been easy to use and is meeting our management and budget control needs at scale.”

The RONIN platform provides easy and instant access to incredibly powerful Amazon Elastic Compute Cloud (Amazon EC2) instances such as the Amazon C6i Instances powered by the latest Intel Xeon Scalable processors. Using the RONIN platform, researchers can launch HPC environments with AWS ParallelCluster and gain immediate access to compute resources with just a few clicks. “We strive to make everything simpler so more researchers can produce more scientific outcomes,” says Nathan Albrighton, chief executive officer and founder of RONIN. RONIN enables a single UBC researcher to spin up their very own 8,640-core cluster with 17 TB of RAM at peak compute capacity.

After a year-long pilot program, RONIN worked with the ARC team to implement the ARC Cloud Platform in production. “We wanted to get a variety of users and try different workloads on the RONIN platform before moving to production, and the pilot helped us get some solid feedback from researchers,” says Venkat Mahadevan, cloud architect at University of British Columbia’s Advanced Research Computing unit.

The RONIN ARC Cloud Platform is designed to make it easy for UBC researchers to launch and manage HPC resources without needing to lean on IT for help. “The RONIN platform includes an intuitive web interface that simplifies the launching of compute environments for research, without researchers needing to learn how to manage network or security groups,” Albrighton says. Additionally, RONIN integrates Intel tools into the platform at the click of a button for a centralized management experience. “RONIN integrates the Intel Compiler suite into the HPC launch process,” says Mahadevan. “That makes things easier, and in terms of compatibility, the software is validated to run and produce the correct results on the platform.”

Now, UBC researchers with AWS funds or credits can easily spin up clusters containing thousands of cores. “Any researcher across the entire UBC campus can access compute resources on demand, whenever they need to, without waiting in a queue,” Albrighton says. Mahadevan adds, “For one of our computer science researchers, RONIN allows their students to use the solution without learning complex technical features, and thus focus on their project goals without unnecessary IT distractions. This has been a challenge in the past, and RONIN reduces the complexity of HPC computing through the RONIN cloud platform.”

Although the majority of UBC HPC research workloads still run in an onpremises or national computing environments, RONIN provides the additional cloud-based compute capacity researchers often require. “RONIN gives us a just-in-time resource for additional capacity that helps researchers meet deadlines or scale up in a way they can’t always do on premises,” says Mahadevan. UBC researchers taking advantage of RONIN can easily access as much compute capacity as they need to supplement the finite fixed resources in the on-premises solution. “RONIN gives us an option that has significantly raised the ceiling when it comes to scalability,” Smith says.

With no technical limits on their ability to run HPC workloads, UBC researchers are completing new research in multiple areas. For example, economics researchers are using RONIN to power use cases with Matlab, a widely used environment for programming and data analysis. Other researchers are using the platform to run complex computational fluid dynamics models, while another university professor is using RONIN to build and train models for Arabic natural language processing.

UBC researchers can also use RONIN to better manage HPC costs. “Our research community gets a certain amount of funding, which has to be weighed against purchases for computational resources,” says Smith. “RONIN provides predictions for what budget cluster setups will consume in the long term, which gives researchers confidence and better budget oversight.”

UBC ARC now plans to expand the use of RONIN throughout the UBC campus. “We launched RONIN as a service fairly recently, and we’re about to initiate an outreach program to bring more visibility to what we have here at ARC,” Smith concludes. “Research is always at the center of everything we do, and RONIN and AWS are among the solutions we use that help us support that goal.”