Berkeley Research Computing (BRC) offered two training sessions on how to use Savio, the campus’s high-performance computing (HPC) cluster, in late summer and early fall 2016. Both the introductory session on August 2nd, and the intermediate session on September 27th, drew sizeable and enthusiastic audiences to the Academic Innovation Studio in Dwinelle Hall. Chris Paciorek of the Statistical Computing Facility and BRC presented both sessions, with backup from members of BRC’s HPC and Consulting teams.
More than forty campus researchers attended the introductory training session. Attendees included graduate students and postdocs, as well as six undergraduate researchers. A significant fraction of these represented the biosciences and public health domains. Topics covered included logging in to the cluster, accessing software, running jobs, and transferring data files. During this session, researchers asked dozens of widely-ranging questions about logins, software modules, running jobs across multiple cores, storage options, and more.
The intermediate training session focused on software installation and parallel computing, and drew 25 participants. Examples of questions asked by researchers at this session included how Python treats local installation of its packages within a user’s home directory; the distinction between process threads and parallel processes; and determining how many cores one’s computational jobs will require, along with how to choose among the multiple SLURM scheduler options for requesting use of those cores.
Materials from these training sessions have been posted on the Research IT website’s HPC Training and Tutorials page. If you have requests or suggestions for future training sessions, or if you’ve come across examples of particularly effective in-person or online HPC training or tutorials to which you’d like to direct the BRC team’s attention, you are cordially invited to contact us!