Abstract: An application server maintains a first plurality of applications in non-volatile memory. The application server loads into volatile memory a subset of the first plurality of applications. The subset is a second plurality of applications. The application server receives a request to execute an application that is not currently loaded in volatile memory, and in response performs a set of operations: (1) determining that there is inadequate storage space in volatile memory to store the application; (2) identifying at least one application in the second plurality of applications that is designated as exempt from eviction from volatile memory; (3) selecting from among the second plurality of applications an eviction target, where the selection process excludes the designed applications; (4) evicting from volatile memory the eviction target; and (5) loading and executing the application in response to the received request, and returning a result responsive to the received request.