Abstract: Presented and implemented in this paper is a scalable request routing mechanism for supporting scheduling policy based on request content, which is central scheduling and distributed routing Unlike common single front end, the scalable routing architecture is composed of several front ends One front end performs the central scheduling task, i e the dispatcher, while the others carry out distributed routing tasks, i e distributors The testing results show that the throughput of a PⅢ 933 host on which the scheduler runs is 450000 conns/s, capable of satisfying 80 distributors requesting scheduling service concurrently In other words, 80 distributors can distribute request or response at the same time, i e distributed front ends can provide throughput up to 16 Gbps