Remember to frequently read the ../readme.txt or ../../readme.txt file (if present), respective of your storage location, for updated terms of usage.
Do NOTE that the storage you get is NOT backed up, and we will periodically ask that you clean your files to accommodate new students/employees.
The DTU HPC infrastructure is composed of sub-systems which all work together to provide an efficient and performant platform suitable for a wide range of workflows. For the uninitiated, one might suspect that storage is just storage. However, this is not the case when dealing with high performance computations. Therefore, when asking for additional disk storage at the DTU HPC, we ask that you provide as much information as possible so we can:
a) choose the right storage size, and
b) choose the optimal partition for best performance.
Please answer the following questions when contacting us for requesting additional storage (a template text is found at the end).
1
Program(s)?
Please tell us which programs you plan to use.
Choosing the optimal storage location depends on the programs because they differ in how they access the storage space.
Some use persistent and few files, while others are frequent visitors of many smaller files.
2
GPU’s
Will you be using GPU’s in your workflows?
A GPU may require frequent read/writes to lots of data-files and/or many random-access to files. For such cases special storage pools capable of feeding the GPU with high enough bandwidth becomes important for sufficient utilization of the GPU.
If you know the access patterns to files, please share how the GPU interacts with the file-system.
3
Student?
Are you a student, and is your activity part of a course?
Please let us know the course # (find it here) so we can best assist you and your fellow students.
Also, if your usage is part of an BSc/MSc thesis, please share with us the supervisors details.
If you have the contact details of the teacher who requested that you used the HPC infrastructure, it will help us faster assign the resources for the course activity.
4
Space required?
We kindly ask that you critically consider the amount of space you request.
Our default storage option is 100 GB.
There are various pools of storage options that may perform much better under certain conditions, it is vital to understand the needs before choosing the best option.
- Will the project produce or require a large amount (say >1000) of small (temporary) files when executing your program? Or will it primarily produce fewer, larger files?
- How much of the requested space will be designated to a dataset, i.e. persistent data?
- How much of the requested space will be used consistently across the project?
5
Lifetime?
What is the lifetime of the data/project?
Is there an end-date of the project so that we can clean up the data?
Here is a template snippet you can copy-paste into your reply mail to the support team:
1. Program(s)?
2. GPU's?
3. Student/Course?
4. Storage size?
5. Lifetime?
Other relevant information?