Dynamic application layout is the capability to move the components of a distributed program among different hosts during the execution of the application. This capability is essential for large-scale applications since it allows to adapt them to changes in resource availability, which are common in wide-area environments. The FarGo system introduces a model for programming the layout of distributed applications separately from their basic logic, by attaching relocation semantics to inter-component references, and by using a built-in monitoring support for making relocation decisions. Dynamic layout policies are encoded within the application using a special API or externally using a high-level scripting language. This paper presents the design of the runtime environment that realizes the model.