Jupyter Notebook Service
The Jupyter Notebook Service, powered by JupyterHub, provides an interactive computing environment at no additional cost to users of the basic service (including supercomputer users, excluding use by private companies and similar entities). This service is provided on a best-effort basis and is subject to limitations such as the number of simultaneous connections.
- Jupyter Notebook Service: https://potato.hucc.hokudai.ac.jp/
Computing Resources
- CPU profile: 2 virtual cores, 4 GB memory
- GPU profile: 8 virtual cores, 16 GB memory, 1 MIG* instance (1g.10gb)
*MIG (Multi-Instance GPU): A feature of NVIDIA GPUs that allows hardware partitioning into up to seven separate instances.
Storage
-
100 GB per user (separate from cloud storage and supercomputer storage; capacity cannot be increased)
Using package management tools such as pip, users can install any Python packages of any versions they need in their personal environment (the environment is retained even after logging out).
Available Profiles
- CPU profiles:
- minimal-notebook
- datascience-notebook
- all-spark-notebook
- GPU profiles:
- tensorflow-notebook
- pytorch-notebook
Usage Limitations
- Maximum simultaneous connections: CPU Profiles: 200+ users; GPU Profiles: 14 users
- Maximum computation time per session: 6 hours
- Maximum idle time: 30 minutes
The allocated resources for the Jupyter Notebook Service may be reviewed and adjusted as needed anytime to balance with resources for the Kubernetes Cluster Services.