ON/OFF Model of Conversational Pattern in Instant Messaging System

Along with the rapid progress of Internet technologies, Instant Messaging (IM) has become increasingly popular due to its social functionalities. Despite its prevalence, little has been done to analyze the characteristics of the conversational pattern among IM users. In this paper, we establish an ON/OFF model to simulate single user conversations in IM system. With data collected from popular Chinese IM software QQ, extensive comparisons have been conducted both on statistical and queuing performances. We therefore confirm that the ON and OFF periods are best-fitted with the log-normal distribution. With the mathematical understanding presented by this paper, we can further design or optimize IM systems toward a more resource-efficient direction, and alleviate its potential impact on wireless networks where resources require meticulous allocation.