A data processing network, server device, and method in which an application program memory usage parameter is monitored where the parameter is indicative of the server device's performance and loading. If the memory usage parameter exceeds a specified criteria, the amount of the system memory available to the application program is reduced and a physical section of memory is deactivated to save power. The parameter may represent the server application's file cache hit rate and reducing the amount of memory available to the application program may include reducing the file cache size. Reducing the file cache size may include invalidating file cache data based upon a purge criteria that indicates when the data was most recently accessed. If the memory usage parameter falls below the specified criteria, additional system memory made be activated and made available to the application program to maintain performance at a desired level.