In developing an application comprising various internal microservices on Google Kubernetes Engine (GKE), you seek the ability to configure each microservice with a distinct number of replicas. Additionally, you aim for a uniform approach to address any microservice from another, irrespective of the microservice's scaling. How should you implement this solution?