System requirements
This page describes what infrastructure the OpenLM Platform needs to run, regardless of the deployment path. For path-specific sizing, see the requirements page within each deployment guide path.
What the platform needs
The platform requires a Kubernetes cluster and four infrastructure services:
| Component | What it is | Why the platform needs it |
|---|---|---|
| Kubernetes cluster | Container orchestration | Runs all ~150 microservices, handles scheduling and self-healing |
| SQL database | Either a single MSSQL instance, or MariaDB/MySQL + PostgreSQL | Stores identity, operational, and reporting data. See Deployment paths for details on the two configurations. |
| MongoDB | Document database | Stores platform configuration, state, and data for some of the services |
| Kafka | Message broker | Event streaming between microservices |
| Redis | In-memory cache | Caching and session management |
These services can be cloud-managed or self-hosted depending on your deployment path. See Deployment paths for what is available in each environment.
Baseline resource requirements
Across all deployment paths, the platform needs roughly the following resources:
| Resource | Baseline |
|---|---|
| Compute | 8 nodes, 68 vCPUs, 136 GB RAM |
| Storage | 300 GB SSD for data services, 50 GB per standard node |
| Kubernetes | A supported distribution with ingress and TLS (EKS for AWS, AKS for Azure, any CNCF-conformant distribution for on-premise) |
| Networking | HTTPS (443) and HTTP (80) ingress for agents and users, outbound access to public.ecr.aws/r3q3q2f4 for container images |
These numbers represent a stable baseline for small-to-medium production load. Detailed node layouts and managed service sizing are covered in each deployment path's requirements page within the deployment guide.
Shared requirements
The following requirements apply to all deployment paths.
- Persistent storage classes must be available for stateful workloads (databases, Kafka).
- TLS termination at the ingress layer with a certificate trusted by all agent machines.
- Validate sizing against expected Broker counts, user volume, report volume, and data retention.
:::warning Sizing The preceding resource numbers are baselines for small-to-medium production load. Validate against your expected Broker counts, user volume, report volume, and data retention before deployment. :::