Institute for Digital Research and Education
We will discuss how to effectively run jobs on Hoffman2 cluster which serves a wide spectrum of user applications and is almost always heavily loaded. We will first review the model and policy of Hoffman2’s job scheduling, followed by fundamentals of using its job scheduler, Univa Grid Engine. A number of common use cases will be discussed in details, including: (1) Single-CPU jobs, (2) Multi-CPU, same-node jobs, (3) Multiple-node (MPI-style) jobs, (4) Job arrays and (5) High-priority jobs.