Skip to content

Community Rules

Follow the guidelines and rules on this page when interacting with the ML Cloud in order to be respectful of your fellow users. All ML Cloud account holders must follow a set of good practices. Exercise good conduct to ensure that your activity does not adversely impact the research community with whom you share your resources or the administrators who maintain and provide your resources.

We have developed the following guidelines to improve and ensure the stability of the ML Cloud. Please familiarize yourself with all the following guidelines.

Login Nodes Best Practices

Login nodes are shared amongst all users. Dozens of users may be logged on at one time accessing the shared systems, including file systems. A single user running computationally expensive or disk intensive operations on a login node will negatively impact performance for other users.

Running compute tasks on the login nodes is an express route to a complete ML Cloud account suspension. Instead, run jobs on the compute nodes using Slurm via an interactive session or by submitting a batch job.

The login nodes are a prep area for your research computational tasks. On the login nodes, users may perform non-computationally-intensive tasks such as editing and managing files, compiling basic code, submitting new and managing existing batch jobs, etc. The login nodes are an interface to the "back-end" compute nodes where the compute work must be performed.

Instead, run your compute tasks on the compute nodes through submitting your jobs to slurm. The compute nodes are where actual computations occur for your research. Hundreds of jobs may be running on all compute nodes, with hundreds more queued up to run when resources are available. Compute nodes are powerful resources, but are also shared among all users. Therefore, there is a limit to how many simultaneously-running Slurm jobs a user may have, with the rest remaining in the queue.

Important

Never run jobs or perform intensive computational activity on the login nodes. Your account may be immediately suspended if your programs are impacting other users.

Job Submission Tips

  1. Request Only the Resources You Need. Requested resources count against your user and group resource-allocation fairshare. Make sure your job scripts request only the resources that are needed for that job. Requesting too many resources may result in your job running inefficiently and consuming more of your fairshare than you need to.

  2. Respect and understand resource limits, especially for GPUs. If you understand the resource limitations of the hardware, your jobs will be more effective. Different nodes have different GPUs, and different GPUs have different amounts of memory. If your application needs more GPU memory than is available, your job will fail and may leave nodes in unusable states. This type of job failure can also cause other jobs on the same node to fail. In some circumstances, this may cause the ML Cloud administrators to cancel your jobs.

Web Crawlers Caution

While web-crawling could be used for legitimate researcher purposes, it may negatively impact ML Cloud operations through i.a. public-IP banning or storage system impairment. If you need to perform such crawler activities, please file a support ticket first. The ML Cloud Team reserves the right to instantly stop or block such crawlers if run without prior written authorization from the ML Cloud Team.

User Responsibilities for the Use of the ML Cloud Account and Services

Don't:

  1. Neglect your security responsibilities; you are personally and professionally responsible for actions taken on your account.
  2. Leave your credentials unprotected for others to access. If your credentials or those of another user have been compromised, immediately notify the ML Cloud support team.
  3. Share your account access with a different person. Each ML Cloud account is assigned to the respective user only. Sharing access credentials will result in an immediate ban from all ML Cloud services.
  4. Interfere with the use of any services by other users or compromise the privacy or security of other users.
  5. Use storage or computational resources for purposes not related to the research project(s).
  6. Never infringe upon someone else's copyright. It is a violation of law to participate in copyright infringement.
  7. Send or transmit harassing, abusive, libelous, obscene, or unsolicited (spam) communications or distribute malicious content.
  8. Tamper with or deliberately disrupt ML Cloud system resources or network traffic to ML Cloud services.
  9. Attempt to breach, circumvent, compromise, or disable any administrative or security controls. Never deliberately scan or probe any information resource without prior written authorization from the ML Cloud team.
  10. Impersonate another person or their credentials in communications with the ML Cloud Team or with external entities.
  11. Engage in any activity which is illegal under local, national, maritime, or international law.

Security gaps that become known to the user or information about apparently erroneously accessible data must be reported to the ML Cloud team immediately.

Last update: September 9, 2024
Created: September 9, 2024