ATTENDING

  • Owkes, Mark <mark.owkes@montana.edu>
  • Johnson, Erick <erick.johnson@montana.edu>
  • Poulter, Benjamin <benjamin.poulter@montana.edu>
  • Young, Mark <myoung@montana.edu>
  • Dlakic, Mensur <mdlakic@montana.edu>
  • Lawrence, Martin <lawrence@chemistry.montana.edu>
  • Wright, Michael <mwright@montana.edu>
  • Rossmann, Doralyn <doralyn@montana.edu>
  • Jerry Sheehan, Pol Llovet, Aurelien Mazurie, Thomas Heetderks

ABSENT

  • Yunes, Nicolas <nicolas.yunes@montana.edu>
  • Sheppard, John <john.sheppard@montana.edu>

MINUTES: Introduction to Hyalite & HPCAG

  1. Overview of HPCAG - Jerry presenting
  2. Background of the cluster - Jerry presenting
  3. The Cluster Queue Structure - Pol presenting
    • discussion: is the ADMIN queue wasteful? It is a tradeoff, it removes ADMIN load from the HEAD NODE. This may no longer be needed now that we have a HIGH AVAILABILITY HEAD NODE.
    • discussion: what if the queue gets overloaded? This will probabably not happen, but if it does, we will make allocations available (we can manually override)
  4. System Administration - Jerry presenting
  5. Cluster Metrics - Jerry/Pol presenting
  6. Compute Hardware - Pol/Jerry presenting
    • discussion: how durable is the storage? All storage is RAID5, very durable.
  7. Cluster Expansion - Pol presenting
    • commented: since we have limited datacenter space, inplace swapping may get us the most in upgrades.
    • commented: would like to see more about educational use

ACTIONS

  • RCi will plan on 70/30 for storage/compute
  • RCi will persue an agreement with UI for disaster recovery backup storage
  • RCi is working with BIOSIT to make samba mounts work (for desktop accessability)
  • RCi is working on an interactive GUI for submitting jobs

FUTURE AGENDA

  • Recearch Project Backups
  • Job Scheduling Optimization
  • Cluster Software