In past years, optimal design of water distribution system (WDS) has been done by minimizing economic costs while meeting the pressure requirement at the critical node, so-called least-cost design. In last two decades, studies have moved to multi-objective problems by maximizing the system reliability while minimizing total cost under pressure constraint. System reliability is generally measured by the probability that the stochastic nodal pressures will be higher than allowable minimum pressure limit under variations of system parameters, such as nodal demands and pipe roughness coefficients. Including reliability as one of the objectives brought idea of system uncertainties into the system design process. Increasing reliability, however, mainly results in increased mean pressure at the critical nodes (with the lowest pressure) and doesn't improve the system robustness interpreted by pressure variations. Robust network is favored since it has less variation on its performance against the system uncertainties. Here a new measure, so-called robustness index, is developed and utilized as an objective for a WDS optimal design. The robustness-based design is compared with the reliability-based design using applications to Anytown network design. The Pareto-front optimization results from both design approaches are compared and post-optimization analyses using system hydraulic availability (SHA) are performed to determine the advantages of the robustness-based optimization approach.