Abstract: Identification and classification of network traffic are an essential prerequisite of network traffic, network management, traffic engineering, and other applications. This paper mainly studies five typical Internet applications, including bitTorrent, HTTP, PPStream, QQ, and thunder, which occupy a majority of the Internet traffic. Analysis of data collected during different time periods shows that each application has a unique typical packet size distribution, and there are only minimal changes in packet size distribution among different durations. Moreover, the analysis on data of discrete sample proportions shows that with decreasing the number of samples, the packet size distribution patterns do not change significantly. With the decreasing of the total number of samples, the packet size distributions curve varies, but the overall shape and the trends have not changed substantially.