Advances in wireless technologies and the pervasive influence of multi-homed devices have significantly changed the way people use the Internet. These changes of user behavior and the evolution of multi-homing technologies have brought a huge impact to today's network study and provided new opportunities to improve mobile data transport. In this thesis, we investigate challenges related to human mobility, with emphases on network performance at both system level and user level. More specifically, we seek to answer the following two questions: 1) How to model user mobility in the networks and use the model for network provisioning? 2) Is it possible to utilize network diversity to provide robust data transport in wireless environments? We first study user mobility in a large scale wireless network. We propose a mixed queueing model of mobility and show that this model can accurately predict both system-level and user-level performance metrics. Furthermore, we demonstrate how this model can be used for network dimensioning. Secondly, with the increasing demand of multi-homed devices that interact with heterogeneous networks such as WiFi and cellular 3G/4G, we explore how to leverage this path diversity to assist data transport. We investigate the technique of multi-path TCP (MPTCP) and evaluate how MPTCP performs in the wild through extensive measurements in various wireless environments using WiFi and cellular 3G/4G simultaneously. We study the download latencies of MPTCP when using different congestion controllers and number of paths under various traffic loads and over different cellular carriers. We further study the impact of short flows on MPTCP by modeling MPTCP's delay startup mechanism of additional flows. As flow sizes increase, we observe that traffic in cellular networks exhibits large and varying packet round trip times, called bufferbloat. We analyze the phenomenon of bufferbloat, and illustrate how bufferbloat can result in numerous MPTCP performance issues. We provide an effective solution to mitigate the performance degradation. Finally, as popular content is replicated at multiple locations, we develop mechanisms that take advantage of this source diversity along with path diversity to provide robust mobile data transport. We demonstrate this in the context of online video streaming, because of its popularity and significant contribution to Internet traffic. We therefore propose MSPlayer, a client-based solution for online video streaming that adjusts network traffic distribution over each path to network dynamics. MSPlayer bypasses the deployment limitations of MPTCP while maintaining the benefits of path diversity, and exploits different content sources simultaneously. MSPlayer can significantly reduce video start-up latency and quickly refill playout buffer for high quality video streaming. We evaluate MSPlayer's performance through YouTube.