Services, like pods, are REST objects. An internal LOAD balancer is PROVIDED by a Kubernetes service. It identifies a collection of replicated pods to which it will proxy connections received. Backing pods can be added or deleted from a service at any time while it stays consistently available, allowing anything that relies on it to refer to it at the same address. The default service clusterIP addresses come from OpenShift Online's internal network and are used to allow pods to communicate with one another.

A replication controller keeps track of how many replicas of a pod are active at any given time. If a pod exits or is removed, the replication controller creates more up to the specified number. Similarly, if there are more running than wanted, it deletes as many as are required to MEET the specified number. With the idea of deployments, OpenShift Online expands support for the software development and DEPLOYMENT lifecycle, building on replication controllers. A deployment just installs a new replication controller and allows it to start up pods in the most basic example. OpenShift Online deployments, on the other hand, allow you to transition from a current image deployment to a new one, as well as define hooks to run before or after the replication controller is created.

When a service is accessed, it is given an IP address and port pair that redirects to the appropriate UNDERLYING pod. A label selector is used by a service to locate all running containers that provide a specific network service on a specific port.