Abstract: According to the empirical study about microblog, the message volume of and releasing intervals of microblog users observer power-law distribution. Through analyzing the relationship between the social network of microblog users and their behaviors, the distribution exponent of message volume is proved to be inversely proplrtionate to the interaction exponent of each user. With the increase of interaction exponent, the power-law exponent decreases much slower. By excluding the influence from other users'forwarding and commenting behavior, it is confirmed that the power-law exponent of the interval distribution is positively correlated with users' interaction exponent. Accordingly, a social-relation-based dynamic model is proposed to reflect the behavior of posting microblogs, such as forwarding, commenting, and social network relation among users. The simulation results of the model match well with empirical data.