Techniques and systems for controlling resources used by a device's web services applications are disclosed. In one technique, web services applications release their allocated resources and enter a “sleep mode” after being inactive for a specified period of time. In one technique, a device maintains a shared pool of memory resources. The pool contains data structures that web services applications are likely to re-use. Before allocating a new instance of a needed data structure from free memory, a web services application first determines whether the shared pool already contains a currently unused but allocated instance of a data structure of the needed type, and, if the shared pool does, then the web service application uses that instance, overwriting existing data in the instance as needed. Web services applications return such data structures to the shared pool when those web services applications are finished using those data structures.