ROAR User Guide   »   Requesting Resources Widely on Roar Collab
Feedback [ + ]

Requesting Resources Widely on Roar Collab

Requesting more resources for a job will increase its time in the queue as it must wait longer for resources to be allocated. Typically, parallelized code will see a reduction in overall runtime as more resources are requested. The total time to solution is ultimately reduced when the job strikes a balance by requesting the minimal amount of resources needed to provide a reasonable speedup. The goal of a resource request is to obtain the necessary computational resources to complete a task while minimizing the time to solution. An optimal resource request is the minimal amount of computational resources that allows a computational task to run to successful completion. It’s useful to examine the amount of resources that a single laptop computer has, or 1 laptop-worth of resources, as a reference. A modern above-average laptop, for example, may have an 8-core processor and 32 GB of RAM. If a computational task can run on a laptop without crashing the device, then there is absolutely no need to submit a resource request larger than this.