RMIT Cuts Research Time from Months to Days with AWS-powered High-Performance Computing

Prior to RACE, RMIT researchers encountered significant delays and limitations in accessing the necessary high-performance (HPC) resources needed for their projects. They were reliant on a number of external, shared facilities for computational infrastructure. Bookings often needed to be scheduled weeks in advance, and access to HPC services was restricted to well-funded projects, significantly hindering the pace of research.

Researchers, attempting to circumvent these delays, sometimes resort to purchasing their own hardware. This often doesn’t meet the specific demands of advanced research projects, comes with significant procurement delays, and results in persistent challenges ranging from hardware that is not fit-for-purpose, to extensive datasets scattered across multiple locations. This approach leads to inefficient research, delayed progress, and suboptimal outcomes.

To resolve these issues, RMIT worked with AWS to launch RACE, a pioneering cloud-first supercomputing facility. RACE democratizes access to advanced computational resources so that researchers can swiftly engage with their projects. This facility has revolutionized RMIT's approach to research, providing robust computing power on demand. RACE fosters a culture of continuous improvement, where a dedicated team constantly enhances and expands supercomputing capabilities to meet emerging research needs.

RMIT University has built an ecosystem of solutions to meet researcher needs. Specifically, RACE provides a catalogue of secure, well-architected cloud environments, and comprehensive training, technical consultancy, and expert services, enhancing researchers’ capabilities and maximizing funding opportunities. With over 550 researchers onboarded, RACE has facilitated rapid testing of ideas and solutions, with some use cases seeing a 100x performance improvement when compared to existing on-premises servers.

The Sir Lawrence Wackett Defense & Aerospace Centre, led by Professor Pier Marzocca, has partnered with RACE to accelerate aerospace simulations. This collaboration has significantly increased the number of analyses that can be run, reducing the time-to-solve by several orders of magnitude. The team is now able to handle 40 million simultaneous equations for more than 400,000 iterations per run, optimizing designs and reducing the number of costly physical tests. The team’s solving time for analyses has reduced from nearly 3 months to just 3.5 days using RACE.

Similarly, Professor Matt Duckham, Director of RMIT’s Information in Society Platform, partnered with RACE to manage, update, and learn from increasing volumes of geospatial data. They built a statewide knowledge graph of Victoria, Australia, with over 2 million interlinked geospatial records processed in under an hour, a task that would typically take days on a regular computer. These partnerships are pushing the boundaries of research and technology, contributing to Australian sovereign capability, and preparing RMIT for the next wave of geospatial data.

RACE utilized several AWS services to advance its research and innovation. A tight coupling of Amazon Elastic Compute Cloud (EC2), Elastic Fabric Adaptor (EFA), and AWS ParallelCluster, an open-source cluster management tool to deploy and manage High Performance Computing (HPC) clusters on AWS, helps streamline computing operations and deliver significant savings in computing resources and time. Amazon Braket, a fully managed service that simplifies access to quantum computers, provides access to an international array of quantum computers. This service allows researchers to focus on prototyping and testing quantum algorithms without worrying about underlying technological complexities. RACE also employs Amazon SageMaker, a fully managed service that supports all phases of machine learning (ML), from data preparation to deployment. Integrated with Amazon SageMaker Studio, this service provides researchers with the tools to develop and deploy artificial intelligence (AI) models efficiently.

By adopting RACE, powered by AWS, RMIT not only sped up research processes but also promoted financial sustainability. With the AWS pay-as-you-go model, the institute can now efficiently manage its research budget, paying only for computing resources used. This model reduces the financial burdens associated with upfront hardware investments and ongoing maintenance costs, and gives them immediate access to the latest hardware without long-term commitments.

The AWS team has been integral to the inception of RACE, working closely with RMIT to tailor solutions that meet its unique needs. Furthermore, this ongoing support has helped to modernize and innovate the University's approaches, transitioning it from traditional methods to cutting-edge, cloud-based research practices.