An important challenge in grids is load balancing, and it includes problems about incentives and selfinterest. Mechanism Design (MD) provides a good solution to these two problems. And a truthful mechanism can force participants to tell their truth if they want to maximize their interest. But a truthful mechanism cannot restrict participants? behaviors in practice. In this paper, we extend mechanisms? definitions, introduce the penalty and bonus, formulate the penalty and bonus model with power functions, and quantify the value with the delay instead of the execution value. Moreover, we apply the model to a specific load-balancing problem, and change the original mechanism with verification into a new mechanism with penalty and bonus. At last we make a simulation study about this new mechanism and prove the effectiveness of the model.