Custom deployments
When to ask for a custom Corvus deployment and what numbers to bring.
A custom deployment is a Corvus resource sized around one workload rather than sold as a preset. Most applications don't need one; the preset plans on the standard regional endpoints are usually the right answer.
When a custom deployment makes sense
| Signal | What it looks like |
|---|---|
| Sustained RPS above the top preset | Steady 429 with -32001 RPS limit exceeded at your normal traffic |
| Weighted-cost workloads | Mostly getProgramAccounts (30 units) or getTokenAccountsByOwner (10 units), so nominal request rate is a fraction of the RPS number |
| Scan concurrency ceiling | 429 with -32005 Scan request timed out waiting for concurrency slot while RPS headroom remains |
| Submission volume | 429 with -32005 Transaction rate limit exceeded |
| Tokenized URL access | You can't pin a stable egress IP; tokenized paths exist on dedicated HTTP and WebSocket endpoints only |
| Tenancy isolation | You want your own resource rather than pooled capacity |
Going custom doesn't widen your choice of region: Solana JSON-RPC, WebSocket and Yellowstone gRPC are offered in Frankfurt, Amsterdam, London, New York and Tokyo either way. See Regions.
Onboarding is authoritative: your endpoint and port, access method and credentials, RPS and TPS, which products are enabled, the region, and where to get support. Where it differs from what you read here, go with onboarding.
Bring your sustained and peak RPS and TPS, and above all the method mix, to us on Discord and we'll size it with you. Weighted methods dominate the real cost, so a rate number without the mix undersizes every scan-heavy workload.