SophiNode uses P2P infrastructure to realize a decentralized E-Commerce application. And service advertisement forwarding is an interesting topic in SophiNode. We discuss speed, scope and overhead feature of a service discovery system, and put forward a Rumor-Spreading model to improve these features. We try to simulate rumor-spreading process of human life, and avoid unnecessary duplicated advertisement overhead. Two key concepts are adopted: Forwarding Probability and Forwarding Weight. The former is in inverse proportion with service advertisement popularity; so that popular advertisements are filtered while rare advertisements are wildly spread. The later helps to determine the Next Hop when forwarding an advertisement. Our simulation shows that after Rumor- Spreading model is adopted, speed and scope feature of service query is improved while overhead feature keeps steady.