When the same resource-intensive data has to be calculated multiple times in parallel programming there is a need for a distributed concurrently accessible central cache. Distributed because performance reasons and it is likely to need more memory than available on a single machine.
This document discusses a framework which provides an easy developing process for memory-intensive applications in grid enviroment. With this framework the whole system memory is accessible concurrently and transparently meanwhile the data consistency is assured.
The framework also supports central cache developing particularly with high-throughput computing systems where minor network latency is granted.
A development process of a central cache based on ordering MCMC simulation algorithm is being presented in this thesis. An average simulation can be run twice as fast as before and with more complex simulations where higher number of variables and concurrent processes are used the results are even more remarkably.