Every node of an engine belongs to one node class, a minimum hardware spec, and the console only offers nodes that match your engine's class and privacy mode.
The five classes, smallest first:
- Nano (NANO): 2 vCPU, 8 GiB RAM, 100 GB NVMe. Hobby projects, learning, staging.
- Standard (STD): 8 vCPU, 32 GiB, 500 GB. Small production SaaS, indie apps.
- Performance (PERF): 16 vCPU, 64 GiB, 1 TB. Mainstream production apps.
- High-Performance (HPERF): 32 vCPU, 128 GiB, 2 TB. Large SaaS, multi-tenant platforms.
- Dense (DENSE): 64 vCPU, 256 GiB, 4 TB. Very large apps, shared hosting engines.
How the class shapes each change:
- Add nodes: the catalog shows only nodes of your engine's class and privacy mode; its Node class and Privacy tech filters are locked.
- Replace: the replacement has the same class and privacy mode as the node it replaces.
- Upgrade class: moves the whole engine to a more powerful class, replacing every node one at a time. Moving down isn't offered, because the period already paid for isn't refunded.
Good to know
- Every node runs the same computations, so an engine moves at the pace of its slower nodes. That is why an engine's nodes share one class.
- SEV-SNP marks nodes with confidential-computing hardware. An engine's nodes are either all SEV-SNP or all Standard privacy, and none of these flows changes that.
- A node's class shows as a badge in its details on Replica nodes.
- Each node's monthly price depends on its class, its provider and its data center; the catalog shows it per row.
- An engine on a class outside these five can't be upgraded; Upgrade class says so.