This paper discusses the implementation using mobile agent and the performance of the market computing to allocate network quality of service most efficiently, based on users' preference. Though the protection of users' private preference and the efficient calculation are both important, these two requirements often contradict each other. By implementing QoS Market, a market-based network resource allocation system, using mobile agents, the communication time between agents and an auctioneer, which is the largest overhead in market computing, is reduced dramatically whthout leeking users' privacy information. The result of experiments show a mobile agent approach is more efficient then others, and the overhead of one market computing cycle in this system is about 450msec, which means that the system achieves sufficient efficiency.