In this paper, a network architecture for fast retrieval of user memory data of RF tag is proposed. The architecture features a network-based resolving of user memory layout in RF tag and an optimization of reading block according to application requests and the characteristics of interrogator. We implemented the proposed network architecture and evaluated the performance with commercial interrogator and commercial RF tags both of which are conformal to the fundamental specification of ISO/IEC 18000-6 Type C. The experimental and analytical evaluation reveals that the network-based memory resolving is effective particularly for interrogators which have large inventory overhead in a reading of user memory data. In our experiment, 3.3 times read speed up is achieved with the network-based resolving. The optimization of reading block is effective regardless of the protocol implementation of interrogator. We achieve 2.5 times speed up in the experiment in three sensor data reading from RF tag which has 2K bits memory.