COMPaaS DLV, COMposable Platform as a Service Instrument for Deep Learning & Visualization, was purchased from Liqid, Inc., with NSF funds. Its 4th-generation computer architecture, composable infrastructure, treats computer components - computer processing units (CPUs), graphical processing units (GPUs), networking, memory and storage - as pools of resources. It is:
Highly flexible. Each application defines the resources it needs, and the operating system composes, or combines, resources on the fly.
Scalable. As more infrastructure is added, it’s auto-integrated with the existing infrastructure and becomes part of the pool of capacity.
High-throughput. System components are interconnected with a high-speed internal fabric, enabling Big Data to quickly move among CPU, GPU, memory and storage at optimum speeds with little to no bottlenecks.
Brown, M., Renambot, L., Johnson, A., Long, L., Visual Data Science: Advanced Cyberinfrastructure Systems (poster session), UIC Foundations of Data Science Institute (an NSF Harnessing the Data Revolution (HDR) TRIPODS grant), Open House, Discovery Partners Institute, Chicago, IL, January 17th, 2020.