Why Kubernetes-native?
Keep cache capacity beside the Pods that use it. When an application gets another replica, that replica gets its own Swytch sidecar and a local client endpoint.
Most caches run on Kubernetes. Swytch is designed for Kubernetes: the cache follows the workload.
A namespaced Swytch resource defines memory, transport, and peer discovery. A Pod-template annotation selects the
profile. The operator injects a native sidecar at Pod creation and creates the discovery Service.
Application Jobs can finish without a lingering sidecar. Existing Pods are not modified in place; changing a profile takes effect when replacement Pods are created.
Use sidecars when each application should have a local endpoint and its own memory budget. Use server mode when several clients need a shared Service or the workload cannot accept a sidecar.
Both shapes use the same Redis interface. Adding replicas does not automatically copy every key to them: subscriptions follow use.
Start with the operator quickstart.