The Knowledge Engineering Group runs its own computing cluster. It consists of 12 nodes with two AMD Opteron(tm) Processor 250 each. One half of the nodes offer 8 GB RAM (6 x 8 GB = 48 GB RAM) while the rest houses 4 GB RAM (4 x 6 GB = 24 GB RAM), providing 72 GB RAM in total.

Debian GNU/Linux squeeze (64 bit version) is installed. The batch system in use is Sun Grid Engine (SGE). In order to monitor the cluster load we have Ganglia up and running. You may use the overview here.

Infrastructure

• slurm-head the controller node, just for submitting jobs
• /nfs/cluster_files is hosted on our Host machine and mounted via automount
• home folder do not work on cluster, home is on each node then only temp dir (and usually empty)
• Environment variables do not work in jobs -> define in job file
• Cluster Nodes: kenode51/52 , kenode43/44 (do not use locally!)
• kenode51(GPU available) and kenode52
  • Nodes: 2x 4 core (HT) nodes with 32 GB RAM.
  • HT (Hyperthreading) means, that it actually shows 8
• kenode44 and kenode45
  • two VMs with each: 2 cores (perhaps also HT), 6 GB
• Cluster Head: slurm-head (in the Stud Subnet)

In case you need computing time for an interesting project don't hesitate to get in contact with us!