Skip to main content

Research Computing announces the availability of an additional special-use partition on the Longleaf cluster to accommodate low I/O jobs such as simulations.  Eventually this partition will be used to alleviate queue congestion and lower Longleaf pend times by bursting compute jobs to the cloud. Over the next few weeks, we will analyze the jobs being scheduled and run locally in this partition to make decisions on how to allocate, and reclaim, cloud compute nodes dynamically. Data locality is a key issue in expanding Longleaf into the cloud, so this new partition is designed to service jobs with low I/O requirements by not having /proj and /pine available. This new partition is available now – start scheduling!

Here are some interesting facts about the new partition:

  • Partition Name: cloudrampBeta
  • 1 hour default job duration and 2 hour max job duration
  • No access to /pine or /proj
  • Full access to /home and /apps
  • One may request up to 250 CPUs per job
  • This partition is in Beta and may change as our analysis progresses