HPC meets interactive Data Science and Machine Learning

Carme (/ˈkɑːrmiː/ KAR-mee; Greek: Κάρμη) is a Jupiter moon, also giving the name for a Cluster of Jupiter moons (the carme group).

or in our case…

an open source framework to manage resources for multiple users running interactive jobs on a Cluster of (GPU) compute nodes.

Follow us on Twitter

→ #OpenCarme

Presentations

(selection)

• Slides from our talk at ISC18 06/2018
• Slides from our talk at LRZ 10/2018
• Slides from ISC 06/2019

Core Idea

We combine established open source ML and DS tools with HPC backends and use therefore

• Singularity containers
• Anaconda environments
• web based GUI frontends e.g. Theia-IDE and JupyterLab
• completely web frontend based
  (OS independent, no installation on user side needed)
• HPC job management and schedulers (SLURM)
• HPC data I/O technologies like Fraunhofer’s BeeGFS
• HPC maintenance and monitoring tools

Job submission scheme

Key Features

• Open source
  • we use only opensource components that allow commercial usage
  • Carme is open source, allowing commercial usage
• Seamless integration with available HPC tools
  • Job scheduling via SLURM
  • Native LDAP support for user authentication
  • Integrate existing distributed file systems like BeeGFS
• Access via web-interface
  • OS independent (only web browser needed)
  • Full user information (running jobs, cluster usage, news / messages)
  • Start/Stop jobs within the web-interface
• Interactive jobs
  • Flexible access to GPUs
  • Access via web driven GUIs like Theia-IDE or JupyterLab
  • Job specific monitoring information in the web-interface
    (GPU/CPU utilization, memory usage, access to TensorBoard)
• Distributed multi-node and/or multi-gpu jobs
  • Easy and intuitive job scheduling
  • Directly use GPI, GPI-Space, MPI, HP-DLF and Horovod within the jobs
• Full control about accounting and resource management
  • Job scheduling according to user specific roles
  • Compute resources are user exclusive
• User maintained, containerized environments
  • Singularity containers
    (runs as normal user, GPU, Ethernet and Infiband support)
  • Anaconda Environments
    (easy updates, project / user specific environments)
  • Built-in matching between GPU driver and ML/DL tools

Roadmap

• since 04/2018: Carme prototype is up and running on our Cluster
• 03/2019: r0.3.0 (first public release)
• 07/2019: r0.4.0
• 11/2019: r0.5.0
• 12/2019: r0.6.0
• 07/2020: r0.7.0
• 11/2020: r0.8.0 (latest)
• 02/2021: r0.9.0 (development)

Documentation

Visit our documentation at doc.open-carme.org.

Who is behind Carme?

Carme is developed at the machine learning group of the Competence Center for High Performance Computing at Fraunhofer ITWM.

NOTE: We are open for contributions!

Contact

→ info@open-carme.org

Sponsors

The development of Carme is financed by research grants from