The virtual device enables seamless use of application services residing on different devices in the vicinity of the user. In a pervasive environment, numerous service combinations can be selected to undertake a task. Current works aim to determine the best possible media services for composition by considering user preferences, environment capabilities and similarity between requested and available services. Previously, the authors considered all of above as well as potential local and remote content sources and destination devices. Here this is extended by considering end-to-end service latency to determine service fitness. The end-to-end delay of a service instance is important to consider as it directly affects the interactivity of the system. Services are selected for composition based on our fitness model. We model and simulate this issue and explain the results of our experimentation.