Carleton is enhancing its fleet of research-ready Graphical Processing Units (GPUs).
Carleton’s Research Computing Services (RCS), which provides researchers with the support, expertise, and resources necessary to conduct research computing at Carleton, has recently acquired four servers housing eight GPUs each.
Two of the new servers were purchased by RCS, while the remaining two were generously donated by Cancer Computer. Altogether, RCS’s offering now includes 42 GPUs across six servers.
Researchers Applying GPUs in Diverse Topics
Carleton researchers across all five faculties are using these GPUs in their work.
“The GPU I am accessing through RCS is a valuable resource that I and my research team are using in natural language processing work on scientific publications data to answer research questions related to the micro foundations of knowledge and innovation,” says Sprott School of Business Prof. Ahmed Doha.
Mechanical and Aerospace Engineering Prof. Reza Kholghy is using a GPU to model the sintering of metallic nanoparticles with carbon impurities using molecular dynamics simulations.
“This research has significant applications in materials science and engineering and can help improve the design and development of new materials, including cathodes in rechargeable batteries, that utilize nanoparticles,” says Hossein Rahbar, a PhD student from Prof. Kholghy’s lab.
Public Policy and Administration Prof. Adegboyega Ojo is using RCS GPUs for qualitative and computational text analytics methods to generate actionable insights from text comments on the National Care Experience Programme.
Computer Science Prof. Olga Baysal is using the GPUs to develop an anomaly detection system based on both deep learning and statistical approaches to monitor the behaviour of system microservices and forecast possible anomalies in the systems’ behaviour.
“RCS has recently supported our research project titled System behaviour monitoring using AI/ML techniques,” says Baysal. “We leveraged the RCS computing resources for setting up the industry-like environment (Kubernetes, Docker), as well as GPU resources for training our deep learning models.”
Prof. Murray Richardson is using the GPUs to process image datasets in the Water and Ice Research Lab within the Department of Geography and Environmental Studies.
“We are using a Linux virtual machine (VM) with a GPU to process massive photogrammetry image datasets collected with drones,” says Richardson. “A typical job will ingest 50 to 100 gigabytes of high resolution camera images to provide three dimensional models of the earth surface in the form of dense point clouds on the order of 109 xyz coordinates. The research computing VM has reduced processing times for these jobs to about 10 per cent of what they would normally take on our desktop workstations.“
Securely Stored and Remotely Deployed
These GPU servers are securely housed in Carleton’s data centres and deployed in its on-premises cloud environment. They are then made available to Carleton researchers remotely by way of Virtual Machines (VMs).
Once a researcher receives access to a GPU server, they can connect to it and get to work using their own computer. This is convenient for both researchers and RCS staff because servers do not need to be physically moved around campus.
This approach, along with the newly increased number of servers, helps to cut wait times for researchers, who are now often able to access GPU servers the same day they request it.
The purchased servers each contain eight NVIDIA Ampere A100 GPUs, adding to a server purchased by RCS in Spring 2022. According to NVIDIA, these GPUs power the world’s highest-performing elastic data centres for AI, data analytics, and high-performance computing.
The remaining servers, containing eight NVIDIA GTX 1080 Ti GPUs, were donated by Cancer Computer. Cancer Computer is powered by passionate volunteers who assist cancer researchers by connecting them with computer hardware, processing capacity and IT support they need to save lives.