You are here: Home / News & Events / DKRZ Tech Talks / Tech Talk: Compute time applications

Tech Talk: Compute time applications

As compute time applications for 2021 are due at the end of October, we provide insights into the process of applying for compute time and will have Robert Sausen, the head of the WLA as a guest for insights into the decision process.
When
Sep 14, 2020 from 03:15 PM to 04:15 PM (Europe/Berlin / UTC200)
Add event to calendar
iCal

Climate scientists from German research institutions can apply for resources at DKRZ. The allocation process for resources starts every year in mid-September, i.e. now, with the opportunity for additional requests in mid-March. Navigating the online forms and writing a proposal can be a daunting task not just for junior scientists. At the same time more than 150 projects apply for resources each year which all need to be processed and reviewed by members of the scientific steering committee. Thus, well-written applications will help the scientists get the resources they need by allowing the reviewers to allocate the resources most appropriately.

DKRZ's Mathis Rosenhauer will show how to apply for resources for a new project and how to request resources after twelve months when you also have to submit a report on your project's progress. While we supply a template for the application document, there are often questions what should go into the proposal and how detailed it should be. These topics will be addressed by Prof. Dr. Robert Sausen of DLR Oberpfaffenhofen who currently chairs DKRZ's scientific steering committee.

Find the full talk here: https://youtu.be/ccTBN1UNHi0

 

The notes of the Q&A session are:

  • Can we group activities from different models and/or related but distinct topics in one “project” for a working group? 

    • Yes. If this is something that fits together, please do so. Make clear that there are different sub-aspects.

  • Should we also mention publications that are being prepared (not published)?

    • Not in the table/ form. Mention your most important results in the report.

    • If you could not use your resources, please explain substantial deviations.

  • what is the typical amount of resources requested (e.g., ~10,000 NH); what range?

    • 5000 to > 1 Mio, it really depends on the project. If you could do your work on a pc/workstation, do it there. // Tell us why you need a supercomputer for your work. Bad code is not a reason...

  • Should we care about the next system?

    • Yes, in terms of changes in the structure. In the future we will have less storage / io per compute power.

    • The new system will have a similar number of nodes as Mistral, so node hours will most likely be “ported” 1:1. Don’t apply explicitly for the new system.

    • The architecture will be similar to the current system. Just more cores per node (128) and more memory. Compilers will stay the same (newer versions /…), but a much smaller change than from Hurricane to Blizzard or from Blizzard to mistral.

  • Testing time to optimize for the new system?

    • You will probably get a chance for free testing while both systems are available.

Document Actions