Understanding fairshare usage for HPCWorks PBS Professional

Configuring fairshare is a great first step but understanding how HPCWorks™ PBS Professional™ tracks usage data is essential for optimizing resource allocation and resolving cluster imbalances. Part of the challenge is that usage accuracy depends on polling intervals and scheduling cycles. Short-duration jobs may not be captured, and pbsfs output can be difficult to interpret.

HPCWorks PBS Professional's fairshare usage database updates throughout each scheduling cycle. Administrators can utilize the pbsfs command to display, adjust and manage entity usage. You can also enhance accuracy by configuring key parameters including fairshare_usage_res, fairshare_decay_time and fairshare_decay_factor. Download this white paper for a detailed look at how fairshare calculates usage and learn how to optimize cluster performance.