cnSGW-C Pod and VM Deployment Layout

This section describes the deployment of cnSGW-C pods and its microservices.

The following figure shows the deployment model of six VMs in cnSGW-C.

VM Deployment Model

In this model, the pods are deployed on VM pairs. Two replicas are available for each protocol pod (for example, rest-ep, protocol-ep, and gtp-ep). One instance is deployed on each protocol VM.

Similarly, service pods and session pods are distributed equally on both the service and session VMs. Such a distribution is controlled by labelling the VMs as well as implementing the K8 affinity and anti-affinity rules during pod scheduling.

This model ensures that, during VM reboot scenarios, at least 50% of the replicas of each pod type are available to handle user signaling.

Graceful pod restart allows pod to complete ongoing processing within 30 seconds. Abrupt pod restart will affect ongoing transactions without impact to PDU sessions.