Overview¶

High-performance computing (HPC) clusters are collections of networked computers and data storage designed for intensive computational tasks. In this guide, we'll explain how to access and use an HPC cluster effectively and responsibly.

HPC Cluster¶

An HPC cluster consists of individual servers called nodes, which are only accessible to researchers remotely. Login nodes serve as gateways to the cluster, allowing you to view files and dispatch jobs to compute nodes, which are specifically configured for computation. A job scheduler manages these tasks, allocating resources and managing executions across the entire cluster.

All compute nodes within a cluster have access to shared filesystems, which is stored on a file server or set of servers connected to a large array of disks. This allows your jobs to access and edit your data from any compute node.

Requesting an User Account¶

To gain access to an HPC cluster, you'll first need to request a credential. All users must adhere to the HPC center's policies. To determine the appropriate cluster for your needs and to request an account, please contact hpc@fieb.org.br.

Being a Good Cluster Citizen¶

When using HPC resources, it's essential to follow best practices and guidelines:

1. Avoid running jobs or computations on login nodes; instead, submit jobs through the Slurm Job Scheduler.
2. Never share your password or SSH key.
3. Do not store any sensitive or critical data on the clusters, except where specifically allowed.
4. Refrain from running a large number of very short (less than a minute) jobs.
5. In addition to these recommendations, the use of the clusters is governed by the official guidelines of CS2I SENAI CIMATEC. Familiarize yourself with these rules to ensure responsible and efficient use of the HPC resources.