Isogeometric analysis (IGA) is particularly suitable for the prediction of buckling load and design optimization of variable-stiffness composite panels , since curvilinear fiber path can be described exactly to improve the analysis efficiency, moreover, analytical sensitivity can be derived to improve the optimization efficiency. In this study, an integrated framework of exact modeling, isogeometric analysis and optimization for variable-stiffness panels is developed for the global optimum. Due to the inherent feature of multiple local optima for this type of problems, a novel multi-start gradient-based strategy is developed to enhance the global optimization capacity , and multiple initial designs for gradient-based optimization are determined by space tailoring method, which can guarantee the convergence rate and efficiency. Once the constraint aggregation and parallel computing methods are employed, the computational efficiency will be further improved. For typical illustrative example, it can be demonstrated that the proposed method is able to provide a more efficient optimum design with significant less computational cost compared to other traditional methods, including FEA-based optimization, direction optimization using genetic algorithm , gradient-based optimization without K–S function, gradient-based optimization based on difference method.