Research Of Anti-Fraud Algorithm In Computational Grid

Computational grid is a vital part of the distributed computations. With the popularization of the personal computer, its performance has been further improved and the internet technology carries more weight in people's life. Computational grid, hence, plays a more important role in the large-scale calculation.For economy or many other reasons, the users of the computational grid will possibly aggrandize or make a false report on its work. Therefore, how to test the correctness of he feedback from the users turn out to be the key point to decide whether the computational grid projects should be reserved in the future. Based on the study of anti-fraud grid algorithm applied in computational grid, the author puts forward his improved edition, and realizes the computational grid application program with globus toolkit. The research is mainly focus as follows:Firstly, the research extends the Non-interactive Commitment-Based Sampling Scheme with such specific function that the users could estimate the computing results without virtually performing the program, which is based on the formula of anti-fraud grid algorithm. Thus, it fully benefits the user by considering the users' franking cost in the computational grid in advance. Meanwhile, the improved algorithm has maintained its accuracy while the number of samples is reduced.Secondly, the research has creatively introduced the incentive and penalty models of economic into the Non-interactive Commitment-Based Sampling Scheme. The introduction of these models into the project encourages the sustaining participation of the users while it also imposes punishment on the dishonest behavior. Also, it eliminates the defect existed in many computational grid projects that the users are all willing to participate the project positively.Finally, the author programmed and examined the improved algorithm under the globus toolkit, making it as a grid service in the WSRF structure, which was released under grid environment. According to the test result from a small sized grid setting, the gird service was proved to be feasible.