Abstract: The common Ecommerce uses the traditional structure, B/S, which provides search engine for users. However, at most of the time, the resources of Ecommerce are distributed and dynamical, which result in many problems such as crowded network and low reliability. To overcome these problems, this paper proposes to move Ecommerce to P2P environment, and replace the search engine with publishing/subscription system. Furthermore, a new filtering algorithm with multithreshold based on contents is integrated into publishing/subscription system to improve the performance and the correct rate of feedback information. The numerical experiment results show that the more personality and robustness can be provided, and the load balance of system can be improved.