Uppsala Multidisciplinary Center for Advanced Computational Science


High-performance computing resources for analyzing sensitive data in Sweden

SNIC-SENS is a SNIC project with the goals:

  • To set up and maintain an eInfrastructure for handling sensitive personal data at the National Genomics Infrastructure, NGI. Resources for this includes:
    • NGI production cluster - Irma
    • NGI production cluster storage - Lupus
    • NGI data delivery - Grus
  • To establish and maintain high-performance and data-intensive computational resources within SNIC for researchers that analyze sensitive personal data from large-scale molecular experiments. Resources for this includes:
    • Research cluster - Bianca
    • Research cluster storage - Castor

Project partners are SNIC, UPPMAX, PDC, NGI, and Science for Life Laboratory.

The project is funded by the Swedish Research Council, the Knut and Alice Wallenberg Foundation, Science for Life Laboratory, and Uppsala University.

Project owner: Ann-Charlotte Sonnhammer, SNIC

Team leader: Peter Ankerstål, UPPMAX

Progress report for the Bianca project 2017-04-07

Part 3 of the project is finished. We have made a system, that sets up project clusters, each one meant to be similar in behaviour to Milou, but with only one research project on each cluster

Today it was put into production.

We have accomplished:

  • /home, /proj, and /proj/nobackup directories are working.
  • Two factor ssh login to project cluster.
  • A "wharf" for copying files between /proj/nobackup and Internet, with two factor authentication.
  • Internet access is otherwise closed. Access between project clusters is not possible.
  • Batch system Slurm is working, with seven GB RAM available for each compute core, with up to 16 cores per compute node.
  • Most of Bianca's compute nodes have 112 GB RAM. Three computes nodes have 496 GB RAM. Five compute nodes have 240 GB RAM.
  • Encrypted backups.
  • Project clusters grow and shrink regarding number of computer nodes, depending on demand and priorities.
  • Each login node uses a single compute core. Missing login nodes are created as soon as a project member logs in.
    Users are automatically logged out after one hour of inactivity.