Scaling and Topology
Grow the deployment without changing how the application connects. Each sidecar remains a local Redis endpoint; keys reach additional nodes through use.
| Stage | What changes | Guide |
|---|---|---|
| One application Pod | One memory budget and a local connection | Quickstart |
| Several Pods on one or more nodes | Peer discovery, subscriptions, aggregate resource needs | Kubernetes scaling |
| Standalone server cluster | Clients connect through a Service | Server mode |
| Multiple clusters or regions | Routable peer addresses and wider shared-key latency | Multi-cluster and multi-region |
Add tiering wherever recovery requirements demand it. Adding topology and adding durability are separate decisions.
The programming model stays the same from one node to a large deployment. Validate the intended workload size with sizing and benchmarks; node count alone is not a capacity guarantee.