When the cache process starts, the Model Repository Service stores each object it reads in memory. When the Model Repository Service gets a request for an object from a client application, the Model Repository Service compares the object in memory with the object in the repository. If the latest version of the object is not in memory, the Model repository updates the cache and then returns the object to the client application that requested the object. When the amount of memory allocated to cache is full, the Model Repository Service deletes the cache for least recently used objects to allocate space for another object.

The Model Repository Service cache process runs as a separate process. The Java Virtual Manager (JVM) that runs the Model Repository Service is not affected by the JVM options you configure for the Model Repository Service cache.