We assume you have read and understood the general OIST policy documents and rules. We link to them below. If you’re ever unsure if or how you should do something, come ask us! We’re happy to help. Also, our main means of contacting you is through your email account registered at OIST. You are responsible for monitoring and receiving any messages there.
Our main means of contacting you is through your OIST email account. You are responsible for monitoring and receiving any messages there.
Your user account is personal. You are responsible for your account, and you are required to protect your account from unauthorized access. Never share your account with others, including other people in your own unit.
You are not allowed to circumvent technical restrictions or work around our rules, and doing so can lead you to permanently lose access to the HPC resources. Even if something is technically possible to do, it may still not be allowed. If you’re unsure if it’s OK to do what you want, please contact us and ask.
Don’t do anything that can disrupt the work of other users, or expose their data. This is true even when you’re doing something that is otherwise allowed.
All computations and data on the HPC systems and storage must be for OIST-related research or training only. Purely personal projects; or any commercial or for-profit use is disallowed. Yes, that includes coin mining. If you’re unsure, contact us beforehand.
You may need to give us attribution in your publications when you are using our resources. You can find out when you need to, and how to do it in our page on attribution.
Do not use restricted resources such as Largemem or Bigmem, or GPU partitions for jobs that don’t actually need those resources. Leave them free for users that rely on them.
Don’t run computations on the login nodes. The login nodes are a shared resource and should only be used for creating and starting jobs, data visualization and building applications. They’re also slow and limited; you’re wasting your time.
Don’t submit thousands of jobs at once, directly or indirectly. Your submitted jobs should never exceed ~2000 jobs on compute or 4000 on short. Also, when submitting multiple jobs (array jobs), each job should run for at least 10 minutes. Very short jobs cause a lot of stress on Slurm.
Don’t run Slurm commands in a loop. This includes running squeue over
and over (in a shell or in a script) to track the progress of your jobs.
Slurm user commands use a lot more resources from the Slurm manager than most
people realize.
We may ask you to stop your job, or change the way you’re running it, in the interest of keeping the systems running smoothly for everyone. We may stop any job at any time if it disrupts the system operation or if necessary for emergency maintenance.
Each unit manages their own directories as they see fit. The unit leader is the owner of, and responsible for all the unit data. Their explicit permission is generally needed to change file ownerships or other access or data management changes.
Don’t store non-research data on Bucket or on the clsuter files systems. This includes personal files, administrative documents, files with personally identifiable information, non-research audio or video files, and backups from other computers.
Illegal and pirated content is absolutely forbidden on SCDA-managed systems, and may lead to a police report, prosecution, and expulsion from OIST.
A shared-use PC is not allowed to connect to Bucket or HPC resources directly if:
For these cases we have a transfer storage system called “HPacquire” that you can use to transfer data. It acts as a security valve: you copy your data to HPacquire from the system, then move it from HPacquire to Bucket or some other storage.
In addition to the HPC policies, you must also understand and agree to all OIST policies concerning use of the IT resources and services. Those policies are listed below.