ALMAS: An Application-Level Memory Management Service

One of the most important concerns that large-scale object-oriented developers are faced with, is to keep the right number of server objects in the main memory in order to respond quickly to client requests. On the other hand, the memory size that these server objects consume should not exceed the upper-bound memory size determined for the server process. Also, the mechanism to keep the right objects in the main memory should keep the hit ratio for the requested objects more than a predefined value, (e.g. 75%). In this paper we introduce ALMAS (Application-Level Memory mAnagement Service), an application-level service that keeps the right number of objects in a defined size of memory to obtain a desired hit-ratio. In this service all of the objects are kept in several object pools, each of which applies a set of policies like eviction policy to its objects. To evaluate this service, we have implemented a prototype in order to compare the performance of ALMAS with the mechanisms are used in a typical operating system. Implementation results show better response time for light-weight objects, managed by this service, compared to the ones which are manipulated by an operating system.