Abstract: In delay tolerant networks (DTN), multi-copies of packet are used to improve performance by many routing algorithms. But uncontrolled redundant packets will increase the payload of networks even infect network communication especially in large scale social networks. The problem is that the redundant packets which can’t be well controlled by routing algorithm will cause the node’s packet buffer overflow and drop useful packets. In mobile adhoc networks (MANET), using cluster architecture is a feasible way to resolve routing problem and decrease the redundant routing packets in the similar scenarios. But the clustering algorithm in MANET can’t be used in DTN directly, because the topology of DTN changes rapidly and the links between nodes connect intermittently due to the high mobility. In social DTN applications, such as vehicles of the public transportation system or campus mobile networks, it can be found that the node’s mobility (such as time and path) follows a special rule. Based on this characteristic, closely social circuit is defined and a clustering algorithm is presented in which nodes with similar mobility rules can be clustered in the same social circuit. A routing algorithm is proposed based on this social circuit architecture, which consists of spray, forward and epidemic phases. It is proved by the experiment that the proposed clustering method and routing algorithm can effectively control redundant packets, and it is more efficient than other DTN routings used in the high payloads and large scale network scenarios.