Welcome to Entropy GPU Cluster’s documentation!

USER'S MANUAL

• Publications
• Hardware
  • Compute nodes
  • GPUs
• Getting started
  • Logging in into the cluster
  • Home directory and disk quota
  • NFS
• Queues and resources
  • Queues (partitions)
  • Quality of service (QOS)
  • User associations
  • GPUMinutes
• Submitting Slurm jobs
  • Time limit
  • Using srun
  • Default Time, CPU and Memory values
  • Using sbatch
  • Environmental variables
• Input data
  • /local_storage_N directories
• Software
  • Singularity containers
  • Software requests
• Jupyter
  • Installing Jupyter software
  • Local SSH setup
  • Cluster account setup
• Registering users
  • Passwordless users
  • Registration form
• Cluster monitoring

Overview

The entropy cluster is a GPU cluster installed at the Faculty of Mathematics, Informatics and Mechanics at the University of Warsaw. Currently, the cluster consists of 11 servers (472 CPU cores and more than 200 000 CUDA cores in 52 GPUs).

Cluster setup

The Entropy cluster setup is based on:

• Ubuntu Server 20.04 LTS.

• Slurm Workload Manager.

• Most parts of the cluster are managed using Ansible.

Access to the cluster

The Entropy cluster is available to the staff and students of the Faculty, but due to the limited resources all usage requests should be sent to professor M. Cygan or professor P. Miłoś.