Video delivery has become a major challenge for mobile networks. The increasing spread of smartphones, along with the expanding availability of LTE coverage, has contributed to an extensively growing mobile video demand, bringing large volumes of video traffic and far exceeding the capacity of mobile networks. As a result, wireless access congestion is becoming more frequent, degrading the Quality of Experience (QoE) for mobile video consumers. In this paper, we present a novel QoE-aware traffic management scheme and architecture for scalable video delivery. Thereby, Scalable Video Coding (SVC) is able to organize video data into layers of different importance, thus facilitating the rate adaptation of video streams. The proposed cross-layer architecture is implemented and the optimization is validated in a real-time streaming prototype and testbed. Results demonstrate the benefit of applying QoE-awareness and cross-layer optimization in congestion scenarios. The perceived video quality is improved considerably for all users in the cell, under both light and heavy congestion.