Over the last four decades, a critical problem in real-time system is to improve the efficiency of the decision algorithm for the rate-monotonic(RM) scheduling. Nowadays researchers extend the decision problem to a generalized optimal design problem, that is, how to adjust the task execution time in the corresponding interval such that(1) the system is schedulable and(2) certain system performance(e.g. CPU utilization) is optimized. All the existing methods for solving this problem are to formulate the problem as the generalized constrained optimization problem(GCOP). However, these methods run very slowly and cannot be applied to the systems with large numbers of tasks. In this paper, a new method for solving the optimization problem is proposed. The method is called tree-like linear programming search(LPS). First, the problem is transformed into a GCOP. Next, the GCOP is partitioned into several linear programming sub-problems. Then, a linear programming search tree is constructed and the node of linear programming is solved by depth-first-search as the optimal solution. The experimental results illustrate that the new method can save 20%~70% of runtime comparing with other existing methods. This work also relates to the research areas of satisfiability modulo theories(SMT), and is expected to improve the efficiency for solving SMT problems.