Page tree

What is Gadi?

Gadi, pronounced ‘gar dee’, is one of the nation’s most powerful supercomputers. A highly parallel system of interconnected computer nodes that supports computationally intensive research. Complete with some of Australia’s fastest filesystems, highest performance research cloud, and visual research platforms, Gadi enables ground-breaking, high-impact research and innovation.

It is pronounced ‘gar dee’, meaning ‘to search for’ in the language of the Ngunnawal people, the traditional owners of the Canberra region, where NCI is based at The Australian University’s Acton Campus.

Gadi was officially launched in November 2019 and replaced Raijin, NCI’s previous super computer in January 2020. The now defunct Raijin supercomputer lives on though, with its processing cores being amalgamated into some of Gadi’s Nodes.

But what exactly is a super computer? And why not use my regular computer?

Supercomputers have been around for decades, the first being the CDC 6600, in 1964. The theory behind current supercomputing is to use thousands of compute nodes to process enormous amounts of data, by allowing them to work on the problem together.

If you’d like, imagine the most powerful computer you’ve ever used, now imagine over 4000 of them all linked together, sharing their computing power. This is what Gadi is, a cluster of over 4000 compute nodes, each of them more powerful than the average desktop computer, all working towards the one goal.

The sheer processing power of Gadi means that large volumes of data can be processed and analysed at a faster rate than ever before. Providing researchers and industries with the computing power they need to thrive.

A 4,962 node supercomputer comprising Intel Sapphire Rapids, Cascade Lake, Skylake and Broadwell CPUs and NVIDIA V100 and DGX A100 GPUs, Gadi supports diverse workloads with well over 10 petaflops of peak performance.

  • 7,200 4-Terabyte hard disks in 120 NetApp disk arrays
  • 20 Petabytes total useable capacity
  • 980 Gigabytes per second maximum performance

Storage systems

We run the fastest filesystems in the Southern Hemisphere, linking high-performance computing (HPC) with high-performance data (HPD) via 100-gigabit network links. NCI's filesystems, catering to the needs of our research community, enable the next generation of computational tasks, including High-Throughput Computing (HTC).

NCI’s filesystems contain around 90 Petabytes of data storage capacity, providing space for research data to be stored in five separate global Lustre filesystems, reaching a total aggregate IO performance of around 450 GB/second.

NCI also runs a special IO Intensive Platform, a dedicated filesystem using 576 2-Terabyte NVMe drives for a cumulative performance around 960 Gigabytes per second.

On top of this, NCI also stores 70+ Petabytes of archival project data in state of the art magnetic tape libraries. In total, NCI runs more than 15,000 hard drives from vendors including NetApp, DDN and HPE.

Our file systems provide data storage for active IO taking place on the Gadi supercomputer.

Cloud Systems

The Nirin cloud provides a high-availability and a high-capacity zone, closely integrated with the Gadi supercomputer and NCI's multi-Petabyte national research data collections.

The cloud comprises a mix of Intel Broadwell and Sandy Bridge processors and NVIDIA K80 GPUs. NCI has repurposed hardware from the previous Raijin supercomputer to provide powerful, new capabilities that enable interactive data analysis and data preparation processes.

Nirin integrates with the Gadi supercomputer and the global filesystems, and supports those aspects of research computing more suited for a rapid-response environment. This distribution of workloads allows for the most productive uses of NCI’s varied infrastructure platforms.

The Nirin cloud platform is tightly integrated with NCI’s supercomputing and high-performance storage infrastructure. Located entirely within the NCI system, this provides privileged, high-speed access to the 70+ Petabyte global filesystems on site. It also makes Nirin the ideal home for data-intensive tools and services such as data analysis environments, Virtual Laboratories and internal data publishing processes.

The name Nirin is a word from the Wiradjuri language meaning 'edge'. This reflects the cloud’s role as one of the computational capabilities at NCI directly accessible to users. The Nirin artwork was created by Anthony Best, a Canberra-based Indigenous artist. The artwork shows the lines of communication meeting at gathering places, represented by the circular features. The arch shapes represent scientific communities researching in the cloud, surrounded by data.