From Documentation
Jump to: navigation, search

Common conversations between researchers and HPC support staff has to do with wait times that jobs have in the queue of the general purpose National Systems like Graham. The wait times that jobs experience in the scheduling queue has to do with several factors, including scheduler configuration, load on the system, account usage, and job resource requests. This webinar follows from previous scheduling webinars and describes how users can examine and interpret scheduler configurations, the current and historical load on the system queue and account usage. The goal is that by understanding these variables that affect wait times in the queue users may modify job resource requests to minimize scheduling delays. Further, this webinar describes known properties of the system that can result in unnecessarily long wait times for some specific types of workloads.