Usage Policy
Davidson College Research Computing group's HPC services are designed to support the college's central missions: teaching and research. Usage of these services must be tied to a project led or supervised by a faculty member. Faculty members are ultimately responsible for all activities carried out by users associated with their HPC project.
Students seeking to use HPC resources for academic purposes who are not enrolled in a course or working with a faculty member may still gain access, but this must be arranged though Research Computing group.
Shared Resource Environment and Usage Guidelines
Because Davidsion College Research Computer Department's HPC systems operate on shared infrastructure, fair access for all users requires adherence to specific usage policies. Violation of these policies, especially repeated infractions, can result in suspension of access.
Access Protocols and Login Restrictions
Access to the Lovelace HPC cluster can occur through a login node or authorized interfaces like web portalals. All compute node access must be managed through the job scheduling system, SLURM.
Login nodes are intended for cluster access and job preparation tasks (e.g., editing code, compiling, submitting jobs). They must not be used for tasks that demand significant compute or memory resources. Such processes must be executed via SLURM, either as batch jobs or interactive sessions on compute notes. GUI applications should be run though HPC remote desktop applications.
Processes that consume high CPU or memory on login nodes may be terminated without notice.
File System Usage and Data Rentention
Scratch storage locations are meant for job-related data and active analysis files only. Attempts to circumvent the automatic purge mechanisms are prohibited.
Efficient Resource Usage and Job Scheduling
Users are expected to request HPC resources accurately and use them efficiently. While precise resource requests (e.g., using the "exclusive" option when jobs need most of a node's memory) may result in longer queue wait times, they improve overall system efficiency.