Containers turned the NAS into an application server. Synology Container Manager, QNAP Container Station, ASUSTOR Portainer, TerraMaster Docker and UGREEN's Docker app all run the same images, and the usual list, a media server, a photo library, a DNS filter, a password manager, a few databases, adds up faster than people expect. This guide lists working memory for common containers, gives totals for typical stacks, and shows what NAS memory each stack needs.
The short answer
| Stack | Container memory | NAS memory |
|---|---|---|
| Light: DNS filter, password manager, a dashboard, a download client | 1GB to 2GB | 8GB |
| Typical: the above plus Plex or Jellyfin, Nextcloud with a database | 4GB to 6GB | 16GB |
| Heavy: the above plus Immich or another photo app with machine learning, Home Assistant, a second database | 8GB to 12GB | 16GB tight, 32GB comfortable |
| Everything plus a VM or two | 12GB plus VM allocations | 32GB to 64GB |
How containers use memory
A container is a process with its own filesystem view, not a separate machine. It shares the NAS kernel, so there is no guest operating system to hold in memory, and it uses only what the application inside it needs at the moment. An idle container costs almost nothing. The exceptions are applications that keep a large cache or run machine learning models, which hold their memory whether or not anyone is using them. That is why a photo library can cost more than five other containers combined.
Working memory by container
| Container | Typical working memory | Notes |
|---|---|---|
| Pi-hole or AdGuard Home | 50MB to 150MB | Trivial |
| Vaultwarden (password manager) | 50MB to 100MB | Trivial |
| Portainer, Homepage, Uptime Kuma | 100MB to 200MB each | Dashboards and monitors |
| qBittorrent, SABnzbd | 200MB to 500MB | Rises with active transfers |
| Nextcloud with MariaDB and Redis | 1GB to 2GB | Database and cache are most of it |
| Plex, Jellyfin or Emby | 500MB idle, 1GB per transcode | Library scans spike higher |
| Home Assistant | 500MB to 1.5GB | Add-ons raise it |
| PostgreSQL or MariaDB (standalone) | 500MB to 2GB | Tunable; performance scales with cache |
| Immich (photo library with ML) | 3GB to 6GB | Machine learning service holds models in memory |
| Paperless-ngx | 1GB to 2GB | OCR jobs spike |
| Gitea or GitLab | 300MB to 4GB | GitLab is the heavy one |
| Frigate (camera NVR with detection) | 1GB to 3GB | Plus a detector; rises with cameras |
Figures are typical steady-state values on a home or small-office NAS, not limits. Check the container's own documentation for anything that matters.
Adding it up
NAS memory = host reserve (4GB) + sum of container working memory + headroom for the file cache (2GB to 4GB). A typical stack of Plex, Nextcloud, Pi-hole, Vaultwarden and a dashboard is about 4GB of containers, which with host and headroom is 10GB to 12GB. A 16GB NAS runs it with the cache intact. Add Immich and Home Assistant and the containers alone are 10GB; that is a 32GB NAS if the model allows it, or 16GB with the understanding that the file cache will be small.
If a VM is on the list as well, add its full allocation; the virtualization RAM guide covers that arithmetic. On ZFS systems add 8GB or more for the ARC.
Which NAS models suit a container stack
Any x86 NAS with a two-slot board and a 32GB maximum handles the typical stack. For the heavy stack, look at models with 64GB maximums: Synology's larger Ryzen plus and xs units, QNAP's TS-x64 and TS-h desktops, UGREEN's DXP6800 and DXP8800. ARM-based entry models run containers but usually have soldered memory and 2GB to 4GB, which limits them to the light stack. The brand guides list maximums by model: Synology, QNAP, ASUSTOR, TerraMaster and UGREEN.
Related guides
How much RAM does a NAS need covers the full range of workloads. Can you run virtual machines on a NAS? covers the heavier path. The memory types guide and ECC vs non-ECC for a NAS cover the module choice, and every listing in the NAS memory collection is built to its model's maximum.
Frequently asked questions
How many Docker containers can a NAS run?
It depends on what they are, not how many. Ten light services fit in 2GB; one photo library with machine learning takes 4GB alone. Add up the working sets in the table and leave room for the NAS itself.
Is 8GB enough for Docker on a NAS?
For a handful of light containers, yes. Once Plex with transcoding, a database and a photo library are on the list, 16GB is the practical floor.
Do containers use less memory than VMs?
Yes, much less. A container shares the NAS kernel and uses only what its application needs. A VM reserves its whole allocation. Ten containers often fit in the memory of one Windows VM.
Does Synology Container Manager need more RAM than Docker on QNAP?
No. Both run the same containers with the same working sets. The difference is in the models each maker allows the app on and in their memory maximums.
Buying NAS memory for one unit or a fleet
Containers are where a 16GB upgrade pays for itself fastest. Request a quote with your models and quantities, or contact our sales team. Related: NAS memory hub, business fleet memory and education and government.