Real-time systems are usually associated with strict performance requirements, especially in safetycritical domains.The bottleneck of performance can be found effectively by stress test.This paper focuses on stress test case generation of task deadlines.And it is expressed as a Constrained Optimization Problem (COP) that includes (1) describing the problem about task scheduling, (2) building the model using Optimization Programming Language (OPL)in CPLEX Optimization Studio, and (3) solving the problem via the CPLEX CP Optimizer.Thus, it can generate stress test cases automatically, which are the sequences of arrival times of tasks triggering worst-case executions.Furthermore, due to the significant increase in memory and time with increasing size of the problem, the uncertainty of tasks arrival time, as well as the insufficiency of a single objective function in some cases, etc., an efficient model is built by changing data structures and constraints, and the validity of the improved model in memory and time is presented via experiment.Aiming at larger-scale problem, a hybrid algorithm preliminary framework is proposed, integrating Constraint Programming (CP) and Genetic Algorithm (GA) together, which can find solutions more effectively and efficiently.