Reference

Last updated on 2026-06-19 | Edit this page

Glossary

apt: The package manager used on Debian-based Linux systems (including Raspberry Pi OS) to install, update, and remove software.
Batch job: A non-interactive unit of work submitted to the scheduler via sbatch. It runs when resources become available and records output to a file.
Cgroup (Control Group): A Linux kernel feature used by SLURM to enforce CPU and memory limits on running jobs, preventing one job from consuming more than its allocated resources.
Cluster: A group of computers (nodes) connected over a network and managed together as a single computing resource.
Compute node: A node in the cluster dedicated to running jobs. Learners do not log in to compute nodes directly; work is submitted via the scheduler.
DHCP (Dynamic Host Configuration Protocol): A network service that automatically assigns IP addresses to devices when they join a network. The login node runs a DHCP server (dnsmasq) for the cluster’s internal network.
dnsmasq: Lightweight software running on the login node that provides both DHCP (IP address assignment) and DNS (hostname resolution) for the cluster’s internal network.
DNS (Domain Name System): The service that translates human-readable hostnames (e.g. node01.local) into IP addresses so computers can route traffic to them.
EESSI (European Environment for Scientific Software Installation): A shared software repository providing pre-built scientific applications for HPC clusters across different processor architectures, accessed via CVMFS.
eth0: The name given to the first wired Ethernet network interface on a Linux system. On the login node, eth0 connects to the internal cluster switch.
Ethernet: Wired network technology using Cat5/Cat6 cables. The cluster’s internal network connects all nodes via an Ethernet switch.
Head node: Another name for the login node — the primary node through which users interact with the cluster.
HPC (High Performance Computing): Computing using clusters of machines to solve problems that require more memory or processing power than a single computer can provide.
hostname: The human-readable name assigned to a computer on a network (e.g. node01). Every node in the cluster must have a unique hostname.
Interactive job: A SLURM job that gives the user a shell directly on a compute node, launched with srun --pty bash. Useful for testing and debugging.
IP address: A numerical identifier (e.g. 192.168.5.101) that uniquely identifies a device on a network. Used for routing traffic between nodes.
Job: A unit of computational work submitted to SLURM for execution on the cluster. Jobs are described by a script specifying resources and commands.
Login node: The node through which users access the cluster via SSH. It runs the SLURM controller, NFS server, and DHCP/DNS server. Also called the head node.
mDNS (Multicast DNS): A protocol allowing devices to advertise and discover hostnames ending in .local on a local network without a central DNS server. Built into macOS and Linux; less reliable on Windows.
MPI (Message Passing Interface): A standard library for writing programs that run in parallel across multiple processors or nodes, communicating by passing messages.
microSD card: The small flash storage card that Raspberry Pis boot from. The OS image is written to it using the Raspberry Pi Imager.
Munge: An authentication daemon that ensures only trusted SLURM daemons can communicate within the cluster. All nodes must share an identical munge.key.
NAT (Network Address Translation): A technique where the login node rewrites outgoing packets so that the compute nodes can reach the internet using the login node’s external IP address.
NFS (Network File System): A protocol for sharing directories over the network. The login node exports /home and /sharedfs via NFS so that all compute nodes see the same files.
nmcli: The command-line tool for NetworkManager on Raspberry Pi OS, used to configure static IP addresses, WiFi connections, and network interface settings.
Node: An individual computer in the cluster. This workshop uses Raspberry Pis as nodes.
NTP (Network Time Protocol): Protocol for synchronising clocks across computers. All cluster nodes must have synchronised clocks for SLURM and logging to work correctly.
OpenMPI: A widely-used open-source implementation of the MPI standard, providing the libraries and tools needed to compile and run parallel programs.
Partition (SLURM): A logical grouping of compute nodes used to organise job scheduling. Not to be confused with a disk partition. This workshop uses a single partition called nodecluster.
PoE (Power over Ethernet): Technology that delivers electrical power through Ethernet cables, allowing a single cable to supply both network connectivity and power to a device.
Raspberry Pi: A low-cost single-board computer (SBC) used as the nodes in this mini-HPC cluster. This workshop targets the Raspberry Pi 4 (2 GB+) or newer.
Raspberry Pi OS: The official Linux distribution for Raspberry Pis, based on Debian. This workshop uses the 64-bit Lite (headless) variant.
sbatch: The SLURM command for submitting a batch job script. Resource requirements are specified with #SBATCH comment directives inside the script.
SBC (Single Board Computer): A complete computer built on a single circuit board, integrating CPU, RAM, and I/O. Raspberry Pis are SBCs.
sinfo: A SLURM command that displays the status of cluster partitions and nodes.
SLURM: The job scheduler and resource manager used in this workshop to submit, queue, and run jobs on the compute nodes.
slurmd: The SLURM daemon running on each compute node. It receives jobs from the controller and executes them.
slurmctld: The SLURM controller daemon running on the login node. It manages the job queue and allocates resources to submitted jobs.
SQL: Not used in this workshop; mentioned only to contrast with SLURM’s simpler accounting configuration.
squeue: A SLURM command that shows the current state of the job queue, including running and pending jobs.
srun: A SLURM command for running a command directly on a compute node. With --pty bash it opens an interactive shell.
SSH (Secure Shell): A protocol for securely logging in to and running commands on remote computers over a network. The primary way learners access the cluster.
sudo: A command prefix that runs the following command with administrator (root) privileges. Required for system configuration changes.
systemctl: The command-line tool for managing systemd services: starting, stopping, restarting, and checking the status of daemons such as slurmd and munge.
systemd: The init system and service manager used by Raspberry Pi OS. It starts services at boot and monitors them during operation.
wlan0: The name given to the first wireless network interface on a Linux system. On the login node, wlan0 connects to the workshop WiFi router, providing internet access and learner connectivity.