The load balancing design implemented in Orbix is based on a distribution of clients across different server replicas rather than a distribution of single requests across server replica.

When a client creates a client-side proxy for the remote object the Orbix runtime contacts the locator for a reference to the requested CORBA server. The Orbix locator chooses one of the replica instances out of its pool of replicas and returns the server details back to the client. So the client ends up with a reference to one particular CORBA server instance. The load balancing is only happening at this stage of establishing a connection to a server instance.

Proxy creation and especially connection establishment is very expensive and time consuming. The overhead of contacting the locator  and creating a new connection to the server each time for each request is in most cases much higher than the costs of executing the request itself. Besides, depending on the number of clients and the frequency of requests for each client it might put a very heavy load on the locator process.