As the mobile Internet gradually penetrates into the daily life, it is becoming the most important way people acquire information. Statistical characteristics of users' data access behavior can be used as the foundation for optimization of network deployment, or even trigger new business models. In this paper, distribution patterns of the subscriber's data access behavior have been discussed both in terms of time and space. In the time, both duration and interval of subscribers' data access follow a lognormal distribution. In the space, the radius of the area where users turn to be active follows a lognormal distribution, as well as the distance between successive base stations. Most users tend to be active in a small area within a short while. These results give us a general view about the data usage pattern in the mobile Internet, which is very useful for the resource planning in the carrier networks.