Service-oriented systems have recently emerged as context-independent component-based systems. Unlike components, services can be created, invoked, composed, and destroyed at run-time. Consequently, all services should have a way of advertising their capabilities to the entities that will use them, and service-oriented modeling should cater for various kinds of service composition. In this paper, we show how services can be formally described by the resource-aware timed behavioral language REMES, which we extend with service-specific information, such as type, capacity, time-to-serve, etc., as well as boolean constraints on inputs, and output guarantees. Assuming a Hoare-triple model of service correctness, we show how to check it by using the strongest postcondition semantics. To provide means for connecting REMES services, we propose a hierarchical language for service composition, which allows for verifying the latters correctness. The approach is applied on an abstracted version of an intelligent shuttle system, for which we also compute resource-efficient behaviors, and energy-time trade-offs, by model-checking the systems underlying Priced Timed Automata semantic representation.