Abstract: In a large scale wireless sensor network, all nodes comprise a peer-to-peer network in the case that the sink node does not exist. Any node has probability to become a data consumer node or data producer node. Sensor network is a kind of Ad-hoc network, meanwhile the energy and computation of an individual sensor node is limited, so complicated protocol design cannot be supported. How to make every random generated consumer node and producer node discover each other and search the useful data is a big research challenge in sensor networks. In this paper we propose a bouncing track based data storage and discovery scheme to solve this problem. It is a data-centric storage approach, which needs the consumer nodes and producer nodes to disseminate their queries or data along their relevant bouncing track. This scheme need not an individual node to keep global information of the whole network. Each node determines how to forward their data only using local information and a predefined reflection angle value. In theory, any two bouncing tracks intersect with each other, which guarantees successful data retrieval. This scheme can also satisfy hop distance bounded search cost when a consumer queries data. It can also guarantee data load balance.