Shared Data Grid Programming Improvements Using Specialized Objects

The shared data programming model is an attractive grid programming alternative to message passing solutions. This paper addresses type related improvements to a shared data grid programming model, by taking advantage of particular interactions patterns and object semantics which result in a broader set of specialized objects. We present analysis aspects of GUN, a grid service layer solution for shared data programming. We start by defining the evaluation criteria in terms of performance, resource and qualitative aspects. A series of experiments are used in order to highlight relevant behavioral and performance aspects of the model's implementation. Next, the analysis of the model is carried out at experimental level where a Java based prototype of the grid universe has been built. Starting with the generic objects and followed by every object type, the object particularities are introduced and the experimental results are described and discussed.