Abstract A major limitation in many structural optimization programs is the use of deterministic constraints on stresses and performance. Overall safety should consider the statistical properties of loads, materials and analysis variables. System reliability analysis has improved as tools for predicting the overall safety under extreme load conditions. This paper presents the optimization of truss structures with system reliability constraints. System bounding as well as Monte Carlo Importance Sampling are compared for accuracy and efficiency. Optimization is done with system reliability constraints imposed on both the performance of the original intact structure (reserve reliability) as well as the structure response after specified accident or damage scenarios (residual reliability). Several truss examples show the benefits of such optimization for selecting both member sizes as well as structure topologies. Different behavior models include ductile and brittle behavior and degrees of strength correlation. The programs allow the designers to specify different accident scenarios and corresponding target reliability levels as well as target reliability for the intact geometry. The efficiency and accuracy of the system reliability calculations as well as the optimization algorithm is discussed.