A service should be scalable. At least two instances should be able to run in parallel, enabling failure tolerance and maintenance downtime. A service can also, later on, scale geographically, be near your customers, and improve apparent performance and application response.

For this scaling to be effective, the application platform will need service discovery and routing, a service that could be used by other services to register themselves and expose their capabilities. Other services could, later on, inquire this service directory and know how to reach these capabilities.

To reduce complexity for other services, a service router can redirect requests to service instances. For example, to send emails, you could have three instances and one central router that would redirect requests in a round-robin manner. If any of those instances go offline, the router will stop redirecting to it and the rest of the application doesn't need to care about it.

Another approach could be to use a DNS approach. The name service is capable of handling registrations to a subdomain, and then, when another service makes simple requests, it will receive one or all of the addresses and connect it as if there was only one service operating: