Mobile communication networks are more and more characterized by the integration of distributed and centralized computing and storage resources. Big data capability thus available throughout such networks will not only deliver enhanced system performance, but also profoundly impact the design and standardization of the next-generation network architecture, protocol stack, signaling procedure, and physical- layer processing. In this article, a mobile network architecture enabled by big data analytics is proposed, which is capable of efficient resource orchestration, content distribution, and radio access network optimization. The protocol stack configuration at each access point and the processing optimization of each layer are presented. Key physical layer designs including reference signals and frame structure are discussed. Moreover, utilizing signals in the transform domains, such as delay, Doppler, and angle, may bring enlarged coherence time of the effective channels. It enables much simpler physical layer design, and effectively bridges the latency gap between big data cloud computing and real-time network optimization.