Surving guide for elcid users

The EXACT-LAB / CNR-IOM DEMOCRITOS cluster, “elcid” for short, is the new HPC facility for CNR people here at Sissa.

how to reach it

The frontend node is reachable as elcid.hpc.sissa.it, username and ssh keys are the same you are using on hg1/quicksilver. if you need an account please see How to get an account

hardware

There are 8 blades available, with 4 AMD CPUs (Opteron 6376) with 16 core each for a total of 64 cores and 128GB of ram for each blade, interconnected via gigabit and infiniband networks.

queue system

The queue system is using torque+maui (like on hg1) There is just only one queue is “blade” (e.g. qsub -q blade)

scientific software

provided as usual via module environment:

  • amd open64 compiler 4.5.2.1
  • stock gnu compiler v4.4.7
  • acml 5.3.1
  • openmpi 1.6.4 compiled against the available gnu and amd compilers

storage:

home filesystem

The home filesystem (/u) has 1.7TB available and it’s cluster-wide visible as shared nfs filesystem.

lustre filesystem

A parallel filesystem (/lustre) is available as scratch area of about 14TB