Load balancing will happen on the client service side. The client service will determine where to send the request and the logic of load balancing will be in the client service. For example, Netflix Eureka (https://github.com/Netflix/eureka) is a service registry. It provides endpoints to register and discover the services.

When the pricing service wants to invoke the demand service, it will connect to the service registry and then find the available services. Then, based on the load balancing logic configured, the pricing service (client) will determine which demand service to request.

The services will then do an intelligent and application-specific load balancing. On the downside, this adds an extra layer of load balancing to every service, which is an overhead.