Sizing and Capacity Planning
Blueprints vary widely in resource needs. Use the guidance below as a starting point and adjust based on the specific blueprint workloads you operate.
Suggested tiers
| Tier | Use case | vCPU | RAM | Storage | Notes |
|---|---|---|---|---|---|
| Dev / Test | Local validation, dry runs | 2-4 | 8-16 GB | 50+ GB SSD | Single service, minimal load |
| Standard | Single blueprint, steady traffic | 8 | 32 GB | 200+ GB SSD | Good baseline for production |
| High Throughput | Multiple services or heavy workloads | 16+ | 64-128 GB | 500+ GB NVMe | Reserve headroom for spikes |
Storage planning
- Allocate space for
data_dir(per-service state) and the manager cache. - Prefer SSD or NVMe for fast artifact download and startup time.
- Budget extra space for logs, metrics, and any blueprint-specific datasets.
Network planning
- Stable HTTP and WebSocket RPC endpoints are required.
- Low latency improves job pickup time and QoS reporting.
- Ensure inbound connectivity for your
OPERATOR_RPC_ADDRESS.
Scaling strategies
- Start with one manager instance per host.
- Scale horizontally by running multiple operators on separate hosts and keystores.
- Prefer the VM sandbox for untrusted or high-risk blueprints (see Sandboxing and Security).