Abstract: A query processing approach based on buffer management is proposed for a mass of periodic queries and ad hoc queries, which are both expected to be supported simultaneously in soft real-time applications in the large-scale sensor networks. With a data storage structure designed in the network, a proper execution region is provided for query processing as a buffer where the nodes are able to map the data of remote nodes and the queries and relevant data both concur, so as to process queries in time. Based on queuing theory, a management model is developed for the queries in the buffer with its optimization scheme designed, thus giving the appropriate width and position of the buffer to design the algorithm of query processing. Test results showed that the approach proposed is efficient to the decrease in query missing rate and energy-saving.