| With the emergence and rapid development of grid technology, the grid is improving all the time. Because grid can support distributed resources mixed together as a whole, more and more resources and virtual institutions take part in grid development. No matter which kind of type the resource belongs to, it can play a part in the integrated grid. Yet the problem grows while resources grow, so how to choose a better resource to accomplish a task and make more users satisfied is becoming the main problem. Grid task scheduling algrithem is the key, and we need a more comprehensive scheduling algorithm to meet user's special requirements. We can get such a algorithm by multi-objective model.In grid scheduling, the users pay more attention to the index of time, cost and reliability of the task. This paper is studying a comprehensive multi-objective grid scheduling algorithm model which is based on N-constraint. The study turns the model to a knapsack problem of multidimensional and more choices. This is a NP problem, so we can only search for the optimal solution. The paper put forward a new multi-objective and heuristic algorithm—NVCA (N-Variable Constraints Algorithm). This algorithm is based on dynamic programming. We add a new algorithm for approximate solution in the process of solution procedure. So we can reduce the solution space and improve solving speed with this algorithm. In the last chapter we prove that NVCA is more effective compared to the algorithm we used now by data analysis.The effectiveness of NVCA can be proved in the chapter three. Because the time data of NVCA in the experiment is generated at random by experience, so we can hardly know the exact time in a new task. This is the reason why we study on the time of task alone in chapter four. We use similarity algorithm thoughts for reference to match the data in the historical data based on CBR. We speculated the time of new task by historic data. We get the results by compared experimental data with efficient algorithm which is existed now. |
