HPC and Cloud Facilities at SURF

Access and consultancy @ UU ITS

UU Research Engineering Team

Goal: Increase research impact with advanced digital technologies


Core activities include:

  • Projects
  • Trainings
  • Consultancy (e.g. HPC)


flowchart LR
  B((RDM Support))
  E --- B --- A
  subgraph UB["UU Library"]
    A(Academic Services)
  end
  subgraph ITS[ITS]
    C(RDMS)
    C --> D(Yoda team)
    C --> E(Research Engineering)
  end

style UB fill:#FFCD00
style ITS fill:#FFCD00
style A fill:#fff9dd
style C fill:#fff9dd
style D fill:#fff9dd
style E fill:#fff9dd

uu.nl/rdm

SURF

Summary

  • For ease of use and plug and play compute: ResearchCloud
  • For very powerful parallel computing: Snellius
  • For structurally processing very large amounts of data: Spider
  • For easy storage and data sharing: Research Drive
  • For cheap longterm storage: Data Archive
  • For cloud solutions on commercial clouds: ResearchCloud & SURFcumulus

Snellius

  • Dutch National supercomputer
  • Cluster for large scale parallel computing
  • Batch job system
  • GPU nodes
  • Fast interconnect between nodes

Specifications

CPU nodes (>1200):

  • 128 CPU cores 256GiB RAM
  • Smallest allocation 32 CPU cores
  • Scratch storage ~ 6 TB

Full Snellius hardware overview: edu.nl/x6jgc

Specifications

Hi-mem CPU nodes (>100)

  • 128 CPU cores 1 TiB RAM (2x4 TiB, 2x8TiB)
  • Scratch storage ~ 6 TB

Full Snellius hardware overview: edu.nl/x6jgc

Specifications

GPU nodes

  • 4 x Nvidia A100 GPU (72 nodes)
  • Smallest allocation 1 GPUs
  • 4 x NVIDIA H100 GPUs (88 nodes)
  • Scratch storage ~ 6 TB

Full Snellius hardware overview: edu.nl/x6jgc

LUMI

  • European supercomputer
  • 379.70 PFlop/s
  • 5th fastest in the world
  • 7th greenest in the world
  • Access via SURF (from April)

SURF ReseachCloud

(VRE)

ResearchCloud

Run virtual machines preconfigured for various research needs:

  • Programming and analysis (e.g. R-Studio, Jupyter, Matlab).
  • Analysis tools (e.g. ASReview, 4CAT).
  • Field-specific software (e.g. GIS).
  • Webservices (e.g. Galaxy).
  • …or simply your own desktop or commandline machine in the cloud.

ResearchCloud

Features:

  • Ease of use: launch preconfigured VMs from the portal.
  • Easy collaboration inside and outside of UU (SRAM platform).
  • Run VMs on either SURF HPC cloud, or on commercial cloud platforms.
  • Special SANE environments for sensitive data.
  • Customizable environments (we can help!).

ResearchCloud

Costs depend on the amount of compute you select and the cloud on which you work. Starts at 25 credits/day (1 CPU), up to 3000…

Our starting budgets are often enough!

ResearchCloud

ResearchCloud

For who is it? Those who…

  • need to be able to collaborate or share with others.
  • need more compute (CPU/GPU), but don’t need access to Snellius or another cluster.
  • have non-traditional HPC needs (analysis tools and services).

ResearchCloud

Some example usecases:

  • Training models
    • …e.g. in combination with GIS
  • Transcribing audio/speech recognition (WhisperX).
  • Complex data workflows with multiple users.
    • including with sensitive data

Funding

Getting started at UU

Contact us via RDM support edu.nl/v7y7e

  • HPC consultancy meeting
  • (some) User support
  • Help with compute application
  • Free credits for Snellius or ResearchCloud
  • Engineering projects

UU credits

  • 10.000 credits for Snellius or Research Cloud
  • ~10.000 CPU core hours or 50-100 GPU hours
  • running tests/benchmarks
  • often enough for finishing a project!
  • low threshold intake procedure
  • Use it to prepare a Small compute application