About the project
The use of one of QRIScloud’s large memory nodes has given the Queensland Brain Institute a competitive edge in the prototyping of new research domain specific information technology facilities.
Jake Carroll (pictured), QBI’s Senior Information Technology Manager (Research), said QBI used the node as a deconvolution pipeline prototype for its high throughput microscopy facilities.
Deconvolution is a computationally-intensive image processing technique that is being increasingly used to improve the contrast and resolution of digital images captured via high-end microscopy facilities.
Mr Carroll said QBI was appreciative of the consolidated compute power in the large memory node and the support that QCIF, Nectar and UQ’s Research Computing Centre provided, saying the large memory node “has given us a significant advantage in the production of publishable outputs as we work towards a mature deconvolution pipeline.”
“The large memory node was instrumental in prototyping what an ‘at scale’ deconvolution facility might look like. Without it, we’d have had to find another dedicated piece of hardware to undertake testing of fit-for-purpose infrastructure,” said Mr Carroll.
More recently with the emergence of new GPU-based technology, QBI has been able to upgrade the pipeline developed on QRIScloud onto specialised hardware. "It [QRIScloud’s large memory node] served its purpose in deciding whether the technology was tenable or not,” said Mr Carroll. “The technology now exists to do what we wanted to all along, speeding up deconvolution about tenfold through the use of nVidia CUDA interfaces on K80 series GPUs.”