Description of the ELvis remote visualization cluster

The ELvis cluster (formerly named the RVS cluster) consists of 9 render nodes and a separate login node, each with the following specs:

  • 2x Intel Xeon CPU E5-2620 v2 @ 2.10GHz (6 cores per CPU)
  • 64 GByte RAM
  • 2x NVidia Geforce GTX 780Ti card, with 3 GByte video RAM
  • OS: Debian Linux 7.4 (64-bit), kernel 3.2.54
  • Roughly 800 GByte local scratch space per node

The login node, elvis.surfsara.nl, is only meant for compiling software, testing, inspecting data before starting visualization and starting interactive visualization jobs. The actual visualization jobs will run on the node s37n1 - s37n9.

The scheduling system uses a number of queues to make sure several users can work on the cluster at the same time, while still allowing almost the full cluster to be used by a single user that needs a large amount of memory, or for very long jobs on a few nodes (like producing animation frames). Note that you don't have to explicitly request a certain queue, the choice will be made depending on the job request.

Queue Max. nodes Max. walltime
short 14 1 hour
average 8 8 hours
medium 4 2 days
long 2 5 days