The possibility of using Convolvulus cneorum L., a native Mediterranean xerophyte, with compact dome-like canopy and extended blooming period, on extensive green roofs in areas with semiarid Mediterranean climate was investigated in a 27-month experimental period, which included three summers (the dry season of the year). The aim was to preserve the local character and biodiversity, as well as to reduce water consumption and construction weight. Convolvulus cneorum rooted cuttings were planted in the beginning of July 2011 in experimental modules on a fully exposed flat roof at the Agricultural University of Athens, with a green roof infrastructure (substrate moisture retention and protection of the insulation, drainage element, and filter sheet). Two types of substrate with 10 cm depth were used, one with soil, i.e., grape marc compost:perlite:soil:pumice (3:3:2:2, v/v) and a lighter one without soil, i.e., grape marc compost:perlite:pumice (3:3:4, v/v). Two irrigation frequencies were applied during the dry periods, i.e., every 5 days (normal) and 7 days (sparse) in 2011 and 2012 and every 4 days (normal) and 6 days (sparse) in 2013. The chemical properties of the two substrates were similar, while their physical properties differ slightly as the substrate that contained soil was holding more water at saturation and it had lower saturated hydraulic conductivity and higher easily available water (EAW). The substrate type affected growth since plant height and diameter, shoot number, and aboveground dry weight were promoted by the soil substrate. Irrigation frequency did not affect plant growth. However, plants cultivated on soil substrate and irrigated normally had the highest growth, particularly compared with plants in soilless substrate under sparse irrigation. Flowering was abundant in April (spring) and in the first year flower number was promoted by the soil substrate. During the dry periods, sparse irrigation resulted in increased stomatal resistance one day before irrigation, indicating that water availability was marginal for the plants, while normal transpiration rate was restored the day after irrigation. According to photosystem II photochemical parameters measured one day before and the morning after an irrigation event, no evidence of damage to the photosynthetic apparatus was recorded in any of the treatments. In general, after 27 months of culture, plant size and roof coverage was appearing more or less similar in all the experimental treatments, therefore the combination of the lighter soilless substrate with sparse irrigation is highly suggested for C. cneorum cultivation on Mediterranean green roofs.