Plan Your Deployment
More than one appliance can be deployed. Before deployment, decide how many appliances are needed and where to place them, based on the network topology and the Model endpoints to be evaluated.
Network Location
Deploy each appliance in a network segment that can reach the model endpoints it needs to evaluate, particularly endpoints served by internal infrastructure rather than a hosted provider. Prefer deploying an additional appliance into an otherwise-unreachable segment over opening firewall rules or adding routes across segments.
To reach a model endpoint from the appliance, confirm:
- Routes between the appliance’s network and the endpoint’s network.
- Firewall rules and ACLs that allow the appliance to reach the endpoint on the port it serves.
This is separate from the appliance’s own connectivity back to LatticeFlow AI — see Network requirements for that.
Evaluation Volume
For a high volume of evaluations, or many models evaluated concurrently, deploy additional appliances to distribute the load and reduce turnaround time.
Geographic Distribution
For organizations spread across multiple sites or regions, deploy a separate appliance per site rather than routing all evaluation traffic through a single, central appliance. This avoids:
- Relying on a single network link between sites.
- Added latency when evaluating endpoints far from the appliance.
Each appliance registers and is licensed independently from the Cloud Console — see Register the appliance.