Web Services are built on service-oriented architecture which is based on the notion of building applications by discovering and orchestrating services available on the web. In order to make an orchestrated web service, which is a composition of two or more services, resilient to transient faults, we investigate in this paper the idea of applying checkpointing to web services. We propose a checkpointing policy for generating global checkpoints in an orchestrated web service so that even in case of failure, the composite service can be recovered back to its latest global checkpoint reducing the delay in giving response. We also propose a method for specifying valid set of global checkpoints against which the actual set of global checkpoints generated at run time can be verified for correctness.