Glossary
Key terms used in DS01 documentation.
Container Terms
Container A lightweight, isolated environment that runs your code with its own libraries and filesystem. Containers are temporary - they can be created and destroyed easily.
Image A read-only template (blueprint) used to create containers. Contains the OS, packages, and configuration. Images are permanent and stored on disk.
Dockerfile
A text file with instructions for building a Docker image. Located at ~/dockerfiles/<project>.Dockerfile.
Docker The platform that runs containers. DS01 uses Docker with additional management layers.
DS01 Terms
Deploy
Create and start a container with GPU allocation. Command: container-deploy
Retire
Stop and remove a container, freeing the GPU. Command: container-retire
Workspace
Your persistent storage directory (~/workspace/<project>/). Files here survive container removal.
Tier/Layer DS01 commands are organised in layers:
- L2 (Atomic): Single-purpose commands (
container-create,image-list) - L3 (Orchestrators): Multi-step workflows (
container-deploy,container-retire) - L4 (Wizards): Complete guided experiences (
user-setup,project-init)
GPU Terms
GPU (Graphics Processing Unit) Specialised processor for parallel computing, essential for ML training. DS01 uses NVIDIA data center GPUs.
CUDA NVIDIA's parallel computing platform. Required for GPU-accelerated ML frameworks.
GPU-Slot The unit DS01 allocates. Today (MIG disabled) one slot = one full GPU. If MIG were enabled, a slot could be a single MIG instance.
GPU-Equivalent (gpueq)
The fair-share quota unit — a floating-point compute fraction. A full GPU is 1.0; a MIG instance is compute_slices / 7. With MIG off, gpueq equals the slot count.
MIG (Multi-Instance GPU) NVIDIA technology that partitions a single GPU into isolated instances. Currently disabled on DS01 (the server runs 4 full A100 GPUs); the allocator is MIG-ready should it be enabled.
nvidia-smi Command-line tool for monitoring GPU usage. Run inside containers.
Resource Terms
Allocation
Resources assigned to you (GPUs, containers, memory). Checked via check-limits.
Quota Maximum resources you can use. Configured per-user or per-group.
Idle Timeout
Time after which an idle container is automatically stopped. Typically 30min-2h (varies by user). Run check-limits to see your current value.
Max Runtime
Maximum time a container can run. Typically 24h-72h (varies by user). Run check-limits to see your current value.
HPC Terms
HPC (High-Performance Computing) Using powerful shared computing resources. DS01 is an HPC system for data science.
Fair Share Resource scheduling principle that distributes resources fairly among users.
Cgroups Linux feature that limits and isolates resource usage. DS01 uses cgroups to enforce limits.
File Terms
Host The main DS01 server, outside containers.
Mount
Connecting a host directory to appear inside a container. Your workspace is mounted at /workspace.
Persistent Data that survives container removal. Your workspace is persistent.
Ephemeral Data that's lost when container is removed. Container filesystem (except workspace) is ephemeral.
Network Terms
SSH (Secure Shell) Protocol for secure remote access. How you connect to DS01.
SSH Tunnel Forwarding a port through SSH. Used for accessing Jupyter from your laptop.
Quick Reference
| Term | Meaning |
|---|---|
| Container | Running instance (temporary) |
| Image | Blueprint (permanent) |
| Workspace | Your files (permanent) |
| Deploy | Create + start |
| Retire | Stop + remove |
| GPU | Graphics processor for ML |
| GPU-slot | Allocatable GPU unit (full GPU, or MIG instance) |
| GPU-equivalent | Fair-share GPU quota (float; full GPU = 1.0) |
| MIG | GPU partitioning (currently disabled) |
| Quota | Your resource limits |