Research On Instant Message Network Characteristics

The popularity of instant messaging has changed a lot the way people live. With the increasing Internet users, instant messaging gradually connected the users into a virtual society. As a result the analysis of user behavior and net characteristics of instant messaging is significant.This paper first presents the analysis of QQ protocol and the monitoring of the login process of QQ users. We introduce three methodologies to identify QQ packets. They are port identification, Keep-alive trace identification and Login trace identification. Keep-alive trace identification can identify the QQ packets the most correctly compared to the other two methodologies.Based on the QQ packets sorting by Keep-alive trace identification, we give the statistics on the amount of QQ online users and traffic. They have similar variation trends, which conform to the people's habits of live and net using. However, they don't behave the same way sometime. It is because the lasting time of QQ users chatting is much shorter than that of online. The traffic QQ users spend in keeping alive with servers is main part of total QQ traffic. The statistic on the amount of QQ users login indicates that the frequency of QQ users login reflect the situation of network.Through the mathematical modeling and verification of the way QQ users selecting servers, we find that the probability distribution isn't Bernoulli. The mechanism of server distribution and the design of QQ client software cause the imbalance of servers' load.The statistic on the connectivity degree of QQ user nodes indicates that the network has characteristics of two kinds of networks. The analysis of relationship between server load and network delay also demonstrate that the delay reflect the variation of server load.