The National Computational Infrastructure (NCI) has purchased new storage kit from Fujitsu, NetApp and HPE to support growing demands on the Raijin supercomputer.
NCI specified a large, fast and persistent file system to support bigger data sets for high performance computing with Raijin in 2013.
The machine's current Linux cluster or Lustre storage was deployed in 2011 and has reached the end of its operational life.
The new storage will provide a global file system - gdata1- that doesn't require time-consuming copying of data from one computer to another. Instead, data is accessible and can be shared across all systems.
First, the old Lustre file system devices will be replaced by NetApp E-series storage arrays, provided by Fujitsu. These will also run the Lustre file system.
A second stage replacement for gdata1 will see it move to the Sun Microsystems-developed ZFS file system for Lustre, running on HPE Apollo HPC storage, providing 12PB of capacity.
NCI said the new gdata1 system will use Mellanox EDR Infiniband interconnects with around 70 gigabytes per second bandwidth.
In November last year, NCI settled on Lenovo's NeXtScale cluster with 22,792 Intel Xeon processors, 144 terabytes of memory, and 100 gigabit per second Mellanox EDR Infiniband interconnects to upgrade Raijin's performance.
The NeXtScale cluster will connect to Raijin's new storage systems.
"The additional systems will take NCI’s total data storage capacity to over 36 petabytes, enabling NCI to continue to meet the demand for Australia’s rapidly expanding nationally significant data collections," NCI said.
A $7 million grant from the National Collaborative Research Infrastructure Strategy Agility Fund, which was matched dollar-for-dollar by NCI Collaboration and research institutions, funded the new storage systems.