In this paper, a hardware design methodology for merge-sorting networks, which uses a fixed size Batcher's sorting network, a data memory module and a memory addressing controller, is proposed. In this method, only by adjusting the data flow of the memory addressing controller, the amount of sorting data can be extended easily. Particularly, the adjustment of data flow is quite regular. Therefore, the proposed method has the following merits: low complexity of parallel sorting networks, low hardware fabrication cost, high extensibility, high regularity and no extra data memory space needed. For verifying the proposed approach, a 128-item merge-sorting network has been designed and simulated by Verilog VHDL.