Required cookies

This website uses cookies necessary for its operation in order to provide the user with content and certain functionalities (e.g. language selection). You have no control over the use of these cookies.

Website visitor statistics

We collect visitor statistics on the use of the site. The data is not personally identifiable and is only stored in the Matomo visitor analytics tool managed by CSC.

By accepting visitor statistics, you allow Matomo to use various technologies, such as analytics cookies and web beacons, to collect statistics about your use of the site.

Change your cookie choices and read more about visitor statistics and cookies

CSC

This part of the blog series focuses on the important details of the migration: When things are happening, how they are happening, and how the changes affect service usage, but for those interested in the model itself, the previous blog post offers a deeper dive into the model, the background, and the reasoning behind the changes.

The when and how

To start with the most important question: When is all of this happening? The migration takes place in September 2025, the exact date will be announced later. The goal is to take the new Billing units into use in all services at once, meaning there is no intermediate transition period, where users have to check, which services use new units and which ones use the old ones.

The old unused billing units already granted to projects will be migrated to the new units so nothing is lost. The grants for the new billing units are based on the past usage of the project, so that e.g. projects only running GPU jobs on Puhti don’t end up with Cloud billing units they have no use for. A goal for the migration is that the service usage itself will not be impacted in any way, but resource applications will most likely be unavailable for a time while the old billing units are converted into new ones.

The effects

Once the migration is done, there will be a few new things for users to learn. Firstly, and possibly most importantly, the resource applications will be done in the four new billing units in the future. This means that users have to consider what kind of resources and services they need before applying for them. To support this, CSC will have tools like the service catalog and an updated billing unit calculator to aid users.

Secondly, as usage accounting is also done in the new units, the project reports will also show not only in what services the project has used billing units, but also what units have been used. 

Lastly, the application limits for different resources will be separated based on the capacity available under every billing unit type. So far, all billing unit applications have been processed based on whether they are S, M, or L sized applications, however, an application of 1 million Billing units to be used in cPouta will have a significantly larger impact on the service capacity compared to a same-sized application to be used on Mahti CPU nodes. Thus, the size limits for resource applications will be based on the type of billing unit applied for. The exact limits are yet to be determined and will be announced as soon as they are decided.

New: Billing unit deprecation

As another big change to capacity management, unused Billing units will also deprecate if left unused. In order to improve the predictability of service usage and reduce the amount of unused resources (there are currently enough outstanding billing units to cover usage of all of CSC’s research services for almost 1.5 years), academic projects’ units will be periodically cut if they do not use any. Every three months, counted from the last billing unit grant, the amount of used and unused billing units is checked and, if less than 25% of the available billing units after the last grant have been used, the units are cut accordingly. The goal is to encourage projects to use their resource grants, so no resources are cut from projects using billing units.

As an example, if a project with 40,000 CPU billing units and 25,000 GPU billing units remaining has been granted 60,000 CPU billing units and 135,000 GPU billing units in March, the available billing units after the grant would be 100,000 CPU billing units and 160,000 GPU billing units. This means in June it is checked, if the project has used at least 25% of these (or 25,000 CPU BUs and 40,000 GPU BUs), and if not, the leftover BUs are cut so that 75% of the resources are left. Then in September, it is checked if the project has used 50% of the resources (so 50,000 CPU BUs and 80,000 GPU BUs) and again, if it has not, the resources are cut correspondingly. If the project has used more than the threshold, no resources are cut. If the project is granted new resources, the timer for the resource cuts is reset again.

This cutter is taken into use at the same time as the new billing unit model, so the first round of checks, and possible cuts, is done in December.

These are large changes to CSC’s resource and capacity management, aimed at ensuring that the services CSC offers will continue to keep up with the growing demand and the availability stays on an excellent level. For any questions and more information about these changes, you can turn to the persons listed below, CSC’s service desk at servicedesk@csc.fi , or join the CSC support coffee on 11.6. for a short talk about the renewal and a chance to voice any questions or opinions.

Read more

Headshot

Joonas Nurmi

Project manager

Joonas Nurmi works at CSC as the project manager of the Billing Unit renewal project.

Headshot.

Jussi Tella

Development Manager

Jussi Tella is responsible for support services for resource and customer management.

Headshot.

Atte Sillanpää

Development Manager, Science Support

Atte Sillanpää promotes the usage of CSC computing capacity for research through direct support, documentation and software installations.

+358 9 4572250