HPC Documentation

Guides, references, and tutorials for the WCM cluster

GPU Reservation Policy

GPU Utilization Policy

To ensure fair and efficient use of HPC resources, all GPU reservations must maintain a sustained utilization of 75% or higher. Reservations failing to meet this threshold are subject to review and potential suspension.

Reservation & Performance Guidelines
  • Minimum Duration: GPU reservations are granted for a minimum of two weeks for A100 GPU.
  • Testing Requirement: Users must validate code on a smaller scale before requesting large-scale reservations to ensure the 75% threshold can be met.
  • Monitoring: Usage is monitored regularly; under-utilization may result in a two-week cooling-off period before new reservations are permitted.
Technical Assistance

For help optimizing your code for high-performance throughput or for general inquiries, please contact the support team:

This policy ensures that high-demand GPU resources remain available for active production workloads.