Nodes

Entry point to the cluster where you initially land after logging in.

Lightweight tasks: editing files, preparing and submitting jobs, managing data.

Not intended for running resource-intensive computations.

Long-running or CPU-heavy processes can degrade system performance for all users.

Actual workhorses of the cluster, executing the vast majority of computational tasks.

Accessed via the job scheduler (e.g., sbatch, srun, salloc for Slurm).

Must be allocated through the scheduler to ensure fair resource sharing.

Optimized for heavy computation, memory, or GPU workloads depending on node configuration.

Designed to handle file transfers in and out of the cluster.

Optimized with high-bandwidth connections and specialized configurations to move large datasets efficiently.

Preferred nodes for large or frequent data transfers (e.g., using scp, rsync, or Globus).

Helps avoid overloading login nodes with data copy tasks.