High Performance Computing infrastructure

Information for pilot users of doduo

doduo is the new HPC-UGent Tier-2 cluster, which was installed in October 2020.

Current status & planning

First 64 nodes are online for pilot users.

Planning:

  • maintenance mid Nov'20:
    • add 64 additional nodes
    • downtime for first 64 nodes cabling issues

Technical details

  • operating system: Red Hat Enterprise Linux 8.2 (RHEL8.2)
  • 128 workernodes, each with:
    • two 48-core AMD Epyc 7552 processors (96 cores per node)
    • 250GB usable RAM memory (about 2.5GB/core)
    • 180GB of local disk (SSD)
  • 12,288 cores in total
  • HDR-100 Infiniband interconnect
  • fast access to shared filesystems (GPFS)

Note (Oct 27th 2020): currently only 64 nodes are available, the additional 64 nodes will become available mid November 2020.

Access

doduo is currently only accessible to members of gpilot group.

You can only join the gpilot group by invitation from the HPC-UGent team.

 

To submit jobs to doduo, first run:

module swap cluster/.doduo

Note: the dot ('.') is important!

Software

  • See output of module avail (after swapping to doduo cluster module)
  • Only recent toolchains (foss/2019b, foss/2020a, intel/2019b, intel/2020a)
  • Use the software installation request form if anything is missing, clearly indicate the request applies to doduo!

Resources

Questions or problems

Please contact .