Rocks Cluster Distribution

A specialized distribution for building and managing high-performance computing (HPC) clusters, developed since 2000 by the San Diego Supercomputer Center to make cluster deployment dramatically simpler.

Scientific & EducationscientificHPCcluster-computingacademicTélécharger ↗Site officiel ↗Documentation / Wiki ↗

Basée sur

RHEL/CentOS lineage

Première version

2000

Gestionnaire de paquets

Yum/DNF (.rpm)

Environnements de bureau

None (headless compute-cluster nodes)

Rocks addresses a problem specific to computational science: turning a rack of bare machines into a working compute cluster traditionally required significant manual networking, scheduling-software, and configuration work, repeated for every node. Rocks automates that entire process -- a 'frontend' node installs and configures every 'compute' node over the network automatically, with a modular 'Roll' system for adding cluster software (job schedulers, MPI libraries, monitoring tools) as prebuilt add-on packages rather than manual installs. It's used across research universities and national labs for exactly the use case its name suggests -- getting a functioning HPC cluster running quickly rather than spending days on manual per-node setup.

MirrorNest n'héberge pas les images ISO des distributions — le support d'installation doit toujours provenir directement de la page de téléchargement officielle du projet (lien ci-dessus), afin d'obtenir la somme de contrôle/signature correcte pour vérification.