HPC has come a long way over the past few years, as was evident through discussions and best practices shared at the first Maryland HPC User Group, which took place last week and drew 50 participants from the Washington DC-Baltimore metro area.
Tim Champ from the University of Maryland Baltimore County and Jaime Combariza from Johns Hopkins University MARCC took to the stage first, describing their respective HPC environments and strategies, together totaling over 1300 nodes. Both Tim and Jaime stated that their clusters are managed by Bright Cluster Manager, requiring only 1 FTE each.
Breakout sessions included discussions around roles of storage, networking, and deep learning with HPC. Clearly, high performance data analytics (HPDA) is the next big wave for HPC.
The User Group was organized by Cambridge Computer and hosted at the University of Maryland Baltimore County campus. Other sponsors and presenters were NVIDIA, Bright Computing and Starfish Storage. When it was Bright's turn to present, David Dean and I talked about the Dynamic Data Center, explaining how to manage emerging workloads through a single interface: Hadoop, OpenStack private cloud, and Machine Learning, alongside traditional HPC.
Starfish described how they can manage massive amounts of data from disparate sources through their middleware.
John Olivera from Cambridge Computer did a masterful job in pulling the event together, topping the day off with drinks and dinner. Plans are underway not only for a second Maryland HPC User Group meeting, but to organize similar events in other metro areas, bringing together academic and commercial research organizations.
We look forward to seeing you at the next one!