The technology of multimedia Content Delivery Network (CDN) is to build some replica servers in the network. The origin server replicates some multimedia contents to these replica servers in order to reduce the loading of the origin server and to reduce the utilization of bandwidth. The client issues a request to the most appropriate replica server instead of the origin server, which can provide the same multimedia contents. The main function of the request routing system in CDN is to redirect the client's request to the most appropriate replica server. Some popular request routing algorithms include random choice, round robin, depending on bandwidth availability, depending on hard disk availability, and depending on connection availability. In this paper, we propose a fuzzy decision algorithm using in the request routing system. The fuzzy decision algorithm consists of three steps, which includes fuzzification, rule evaluation, and defuzzification. Based on these three steps, we can get the fuzzy decision value for each replica server. The replica server with the highest fuzzy decision value is the most appropriate replica server to serve the client's request, then the total network utilization will be increased, and the drop rate will be decreased.