-
Roihu dataset projects
Roihu dataset projects
A Roihu dataset project provides a shared filesystem directory for datasets that several computational projects need to use on Roihu. A given dataset project is available directly on Roihu under:
A dataset project provides storage and access management, but no computing
resources. It does not include scratch or projappl directories and cannot
be used to submit jobs.
Dataset project access begins in early August 2026
Applications for dataset projects are already open in MyCSC. The first approved dataset projects will be granted access in early August 2026.
A dataset project is useful when several computational projects need access to the same actively used dataset. Multiple computational projects can be granted read access to the shared dataset directory, while a single owning project has write access.
Using a dataset project helps avoid storing separate copies of the same dataset in multiple Scratch directories. It also provides a clear owner for maintaining the dataset and a central place for managing read access to this data across multiple projects on Roihu.
Dataset projects are granted for one year at a time. Dataset projects are not intended for long-term data storage.
Access model
Each dataset directory has one designated owning project with write access. The owning project is responsible for maintaining the dataset and managing its contents.
Read access can be granted to:
- Individual users
- One or more computational projects
- All users from one or more selected organizations
- All Roihu users (public dataset)
Public dataset projects are readable by all Roihu users. CSC publishes a list of public datasets. Restricted dataset projects are readable only by the users, projects or organisations selected in MyCSC.
Each dataset project is has disk area under dataset, under a specific project ID.
Using data from a dataset project
After you have been granted access, you can use the dataset directly from its directory:
Users with read access can read the files but cannot modify them. If you need to modify data for your own workflow, copy the required files to the Scratch directory of your computational project:
Jobs must be submitted using a computational project:
Do not use the dataset project ID as the Slurm account. A dataset project does not provide computing resources.
Quota and billing
Dataset projects do not receive a default storage quota. The required quota must be requested separately when applying for the project.
Dataset storage consumes Storage Billing Units. Request only the amount of storage that the dataset currently requires.
Dataset projects are primarily intended for actively used datasets up to a few terabytes in size. Contact CSC Service Desk if your dataset is substantially larger or your use case has unusual storage requirements.
Datasets consume the following amount of storage billing units, when compared to other disk areas:
| Storage area | Billing |
|---|---|
| Scratch | 6 BU * TiB * storage-hours |
| Home | 10 BU * TiB * storage-hours |
| Projappl | 10 BU * TiB * storage-hours |
| Dataset — public | 6 BU * TiB * storage-hours |
| Dataset — restricted | 10 BU * TiB * storage-hours |
See more details in the HPC Billing documentation pages.
Applying for a dataset project
You can apply for a dataset project in MyCSC.
A dataset project is valid for one year at a time. An extension must be requested if the dataset is still actively used after that period.
The project application must include an exit strategy explaining what will happen to the data when the project ends. Depending on the use case, the data should be:
- transferred to another storage service
- published or preserved in an appropriate repository
- transferred to the dataset owner
- deleted when it is no longer needed
For detailed application instructions, see Creating a new project: Dataset project.
Dataset projects are not intended for long-term storage and they are not automatically backed up by CSC. Only use dataset projects for actively used data. Maintain a separate copy of any important or irreplaceable data.
For data that is not actively used on Roihu, consider Allas object storage or another storage service provided by CSC or your organisation.