The properties and control of fenestration systems have a significant impact on the energy and comfort performance of buildings. Technology improvement has focused on spectrally-selective solar-optical properties, while careful selection of glazing and shading properties and control allows better balancing between energy and comfort benefits. Despite the advanced modeling capabilities of existing tools, some of these variables are often not considered in the overall design optimization process; therefore, there is no standard way to select optimal window products for specific buildings, orientations, and climates. This study presents a simulation-based optimization method (using coupled whole-building energy simulation coding and genetic algorithm) to select realistic fenestration products in a medium prototype office building, given the uncertainty in shading options at the design stage. A window layer-by-layer optimization approach is followed to minimize site energy use, where various fundamental glass properties (solar and visible transmittance and reflectance, emissivity and conductivity) are the design variables constrained by a comprehensive existing glass library. The optimization is run without shades and with a default shading operation. Each potential solution is a unique combination of all glass properties used as variables. At the post-optimization stage, glazing properties are mapped to real products tagged with product IDs in the database using the minimum Euclidean Distance approach, and complete realistic window systems are generated using the first Quartile optimal solutions. Continuous Daylight Autonomy (cDA) is also calculated to assist in evaluating the overall performance of optimal products under different shading operational modes. The results show the range of optimal window system properties and their performance for three climates (Chicago, Los Angeles, Int. Falls) and two orientations (south and north).