Internet auctions have got attention with the sensational diffusion of the Internet, because we are able to trade various goods for a desirable price. Many Internet auctions adopt the English auction. The English auction is the mechanism to settle a successful bidder. Conventional Internet auction systems are using one logical server called an auction server that is constructed with one or several servers. So, if simultaneous bids occur, the load of the auction server increases. Besides, when congestion at circumference of the auction server occurs, the response time will increase. We propose a new Internet auction method using decentralized selection servers which will reduce superfluous information derived from bidders. The proposed method tends to hold down the increase of the response time by lightening the processing load at an auction server and cutting down congestion around the auction server. In order to evaluate the proposed method concisely, we assume a typical auction service and a network. We evaluate the proposed and the conventional methods by simulation experiments using a network simulator OPNET Modeler. Consequently, it is shown that the proposed method is more scalable and can be deployed with smaller cost.