This page describes what a server needs to run DataMind OS for one organization. Use it when choosing an OS, sizing hardware, planning network access, and configuring firewall rules before deployment.
Linux is highly recommended and is the only platform DataMind tests and supports. Any modern, mainstream Linux distribution with a recent Docker Engine and the Docker Compose v2 plugin works. Ubuntu, Debian, and RHEL/Rocky/Alma are all fine choices.
Running on Windows or macOS (via Docker Desktop) is technically possible for a quick local test, but it is not supported for production use.
| Tier | CPU | RAM | Storage | Fits |
|---|---|---|---|---|
| Minimum (evaluation/pilot) | 2 vCPU | 16 GB | 128 GB SSD | Small datasets, no heavy ETL, evaluating the platform |
| Recommended (production) | 16 vCPU | 64 GB | 512 GB SSD | Everyday production: regular ETL jobs, a normal-sized data warehouse, dozens of BI users |
| Large-scale (heavy production) | 32+ vCPU | 128 GB+ | 2 TB+ SSD | Large data warehouse, many concurrent BI users, parallel ETL jobs, long data/log retention |
We recommend SSD storage. DataMind OS is disk-latency sensitive, especially the data warehouse and reporting components.
There is no hard ceiling above the large-scale tier, and storage in particular is a starting point rather than a fixed number at any tier. Resource and disk needs keep growing with your data volume, user count, and workload. DataMind OS is especially RAM-hungry as workload grows, since the data warehouse, database, and ETL engine all hold much of their working set in memory. If you load large volumes of data, keep years of history, or scale past the table above, check in with your DataMind contact for sizing guidance.
The server needs outbound internet access. Inbound access is not required beyond the ports listed in the next section. Outbound access is used to reach:
| Domain | Purpose |
|---|---|
login. | Microsoft sign-in. The DataMind Installer signs in here with your client secret to download the install files |
unistream. | Container registry. Every DataMind software image is pulled from here |
*. | Registry storage. The container registry delivers image layers from here |
unistream. | Blob storage. Hosts install files (for example, the Docker Compose file) and some UI assets |
platform. | DataMind Platform. Serves the AI features of DataMind OS and the DataMind Installer's Release Notes |
No special bandwidth requirements. A standard internet connection is enough. If your firewall only allows specific outbound domains, allowlist these five.
| Port | Purpose |
|---|---|
| 8000 | DataMind Installer. Needed by whoever performs the initial installation |
| 80 | HTTP (redirects to HTTPS) |
| 443 | HTTPS. DataMind OS itself: the app, dashboards, and reports |
Everything below runs behind the scenes and can optionally be exposed to admin users.
These internal ports should not be exposed to the public internet. Expose them only to your internal network.
| Port | Service | Purpose |
|---|---|---|
| 8081 | Frontend | Direct access to the web app, bypassing the main site |
| 8082 | Backend API | Direct API access |
| 7070 | Carte (Pentaho/Kettle) | ETL tool for building and monitoring data pipelines |
| 8001 | Meltano API | Data connector management |
| 7000 | Rosetta | Document layout and table recognition service |
| 4000 | MCP server | Claude AI tool integration endpoint |
| 9090 | Lurien | Query editor, BI-layout service, direct API access |
| 5432 | Postgres | Main database |
| 8123, 9000 | ClickHouse | Data warehouse (HTTP + native) |
| 8888 | ClickHouse API | JDBC data connector API |
| 6379 | Redis (backend) | Cache and session store |
| 6333, 6334 | Qdrant | Vector database (AI search) |
| 4222, 8222 | NATS | Internal messaging and monitoring |
| 9002, 9003 | RustFS | File storage API and admin console |
| 6901 | Redash | Direct BI dashboard access, bypassing the main site |
| 1080, 1025 | Maildev | Mail testing tool (non-production use) |
| 8002 | Service metrics API | Internal monitoring |
| 9428 | VictoriaLogs | Log storage and viewer |
Linux. It is the only platform DataMind tests and supports. Any modern, mainstream distribution with a recent Docker Engine and the Docker Compose v2 plugin works: Ubuntu, Debian, and RHEL/Rocky/Alma are all fine choices.
For a quick local test, yes, via Docker Desktop. For production, use Linux.
Start from the tier that fits your workload: Minimum for evaluation or a pilot, Recommended for everyday production, Large-scale for heavy production. See Hardware.
We recommend it. DataMind OS is disk-latency sensitive, especially the data warehouse and reporting components.
login., unistream., *.,
unistream. and platform.. If your firewall only allows specific outbound
domains, allowlist these five.
No. A standard internet connection is enough.
443 for DataMind OS itself and 80, which redirects to HTTPS. Port 8000 is needed by whoever
performs the initial installation.
No. Expose them only to your internal network.
Reach out to your DataMind contact before provisioning. It is a quick conversation and saves a resize later.