When a request is received from a client device for the invocation of a web services operation, an off-ramp device coupled to an on-ramp device that provides an interface to web services verifies that the requested service is in a configuration file and then determines the selection scope of a desired service. Based on the selection scope of the desired service, the off-ramp device can determine whether to perform discovery and selection of a service implementation ("request" selection scope), retrieve a service implementation from a session level cache ("session" selection scope), or retrieve a service implementation from an application level cache ("application" selection scope). If a service implementation corresponding to the requested service does not exist in the session or application level cache, the discovery and selection operations may be performed to identify a service implementation to handle the request and the service implementation may be stored in an appropriate cache.