As the performance of computer system improves, the degree of multi-processing grows. Accordingly, need for sharing of resource proportional to priority of process arises for fine grain control of resource sharing.
There are a few proportional sharing resource management algorithms. Some research operating systems are built with proportional sharing resource manangement algorithms. But, there are very little researches about proportional sharing of virtual memory system cause of ambiguity and difficulty with expecting the future memory usage behavior of process.
In this thesis, we introduce the concept that number of page faults occur during a process executes is inversely proportional to the priority of the process to virtual memory system. And, we present proporportional sharing page replacement algorithm based on the concept. We implemented it on Linux kernel and present performance evaluation results of the implementation, and we show that with our algorithm the number of page faults occur during execution of higher priority processes are smaller than relatively lower priority processes and it nearly follows inversely proportional way.
