Optimal task partition and distribution in grid service system with common cause failures

The grid service system is a type of large-scale distributed system, which is being widely used in many fields now. The partition of a grid service task into subtasks and the distribution of them on available resources have great influence on the extent of the service reliability and profits. This paper presents a novel optimization model for maximizing the expected grid service profit and develops a genetic algorithm to solve this optimization problem. As the basis of the objective function, the grid service reliability needs to be modeled and quantified first. However, due to the largeness and complexity of the grid service system, the existing models for small-scale distributed system reliability cannot be directly implemented for the grid. Therefore, this paper presents a virtual model with two-root tree structure that is more general than the star topology. It is because the tree structure not only can represent the virtual architecture of a grid service but also can take into account the common cause failures on the channels that are shared by multiple resources. Finally, a case of a BioGrid application is illustrated for example.