Glossary¶

Amazon Web Services (AWS)¶

A cloud platform provided by Amazon.

Resources: AWS website

Apptainer¶

HPC-focused container runtime. Formerly Singularity.

Resources: Apptainer - Portable, Reproducible Containers , apptainer charm

Canonical Observability Stack (COS)¶

Suite of charms providing open-source monitoring and logging tools.

Resources: Canonical observability documentation

Ceph¶

Distributed storage system that provides object, block, and file storage. Can be deployed and managed using MicroCeph.

Resources: Ceph website

CephFS¶

POSIX-compliant file system interface that runs on top of a Ceph storage cluster.

Resources: Ceph file system , ceph-fs charm

Charm¶

Python software for automating the lifecycle of applications. Also known as a charmed operator. Managed with Juju. Two kinds of charm exist:

Machine charms: charms made to deploy on a bare-metal server, virtual machine, or system container
Kubernetes charms: charms built to deploy on Kubernetes.

Resources: Juju charm definition

Cluster¶

A collection of distinct computers, known as nodes, networked together to act as a single, more powerful system.

filesystem-client¶

A charm that requests and mounts exported filesystems on virtual machines.

Resources: filesystem-client charm

Graphics Processing Unit (GPU)¶

A specialized processor that is designed to accelerate image processing and graphics rendering for output to a display device.

High-Performance Computing (HPC)¶

The practice of aggregating computing power using clusters and parallel processing to complete tasks faster than standard computing.

InfluxDB¶

An open-source, distributed, time series database.

Resources: InfluxData site , influxdb charm

Integration¶

An exchange of data between two charms that allows for interoperability. Formerly known as a relation.

Resources: Juju integrations explained

Job¶

A user-submitted workload managed by the cluster workload manager. Consists of a script containing the executable commands to run an application and declarations of the resources required for that application, such as CPU core count and walltime.

Juju¶

A charmed operator tool that helps deploy, integrate and manage applications across multiple environments.

Resources: Juju documentation

MicroCeph¶

A tool that simplifies deployment and management of Ceph storage both standalone and in a charmed environment using Juju.

Resources: MicroCeph documentation , microceph charm

Microsoft Azure¶

A cloud platform provided by Microsoft.

Resources: Microsoft Azure website

Proxy charm¶

An intermediary charm that enables charms to integrate with non-charmed applications. Also known as an integrator charm.

sackd¶

Slurm Auth and Credential Kiosk daemon. Typically used to provide cluster login nodes.

Resources: Slurm Workload Manager - sackd , sackd charm

Slurm¶

A free and open source workload manager consisting of multiple co-operating software components, each responsible for a piece of cluster functionality.

Resources: Slurm Workload Manager - Documentation

slurmctld¶

Slurm central management/controller daemon. Schedules jobs and monitors other components.

Resources: Slurm Workload Manager - slurmctld , slurmctld charm

slurmd¶

Slurm compute node daemon. Executes jobs scheduled by the controller.

Resources: Slurm Workload Manager - slurmd , slurmd charm

slurmdbd¶

Slurm accounting database daemon. Provides an interface between Slurm and a database for holding historic job statistics.

Resources: Slurm Workload Manager - slurmdbd , slurmdbd charm

slurmrestd¶

REST API interface for Slurm.

Resources: Slurm Workload Manager - slurmrestd , slurmrestd charm

slurmutils¶

Python library for facilitating edits to Slurm configuration files.

Resources: slurmutils GitHub

System Security Services Daemon (sssd)¶

A daemon that manages the retrieval and caching of user credentials and attributes from remote identity providers.

Refers to both the software and the corresponding charm.

Resources: SSSD - System Security Services Daemon - sssd.io , sssd charm

Walltime¶

The maximum duration of a job declared in the job script. Serves as a limit after which the workload manager will forcibly end the job.

Workload manager¶

Software responsible for accepting user jobs, placing them in a queue, and deciding where and when they will run on a cluster - the process known as: job scheduling. Optimizes utilization of a cluster by matching job requirements (CPU core count, walltime) to available hardware while enforcing site-specific usage policies. Slurm is a workload manager.