HPC gives researchers the power and capacity to process large amounts of data (big data) and solve complex computational problems. It is fundamental to the university mission of research and figures prominently in many proposals.
This spring, we migrated our scratch and shared file storage to Weka. This is an all flash-based storage system that provides high performance parallel file access for running jobs and automates tiering to our large object storage array for longer term storage. This summer, we completed adding capacity to our object storage system, which is a shared service between Storrs and Farmington. It provides a total of 5.5 petabytes for the Storrs campus and 2.5 petabytes for Farmington. Through standard lifecycle management, we have leveraged technology advancements to improve the infrastructure. Newer compute nodes offer more memory and faster cores while reducing power consumption. We recently retired 70 nodes in the cluster and added 77. This modest net increase in nodes grew our total core count from 12,500 to 20,148. We also added 38.5 TB of RAM and 25 graphical processing units (GPU), which accelerate and increase data processing power for specific job types.
ITS is committed to the long-term operation and management of the Storrs HPC facility. Planned hardware refreshes are key to maintaining the capital investment, while the associated natural growth in technology allows us to expand both the capacity and capability of the infrastructure to better meet the evolving needs of the UConn research community.