Routing is one of the challenging tasks in Delay Tolerant Networks (DTNs), due to the lack of global knowledge and sporadic contacts between nodes. Most existing studies take a greedy scheme in data forwarding process, i.e., only nodes with higher utility values than current carriers can be selected as relays. They lack an in-depth investigation on the main features of the optimal paths in Epidemic. These features are vital to any forwarding scheme that tends to make a trade-off between packet delivery delay and cost. This is mainly because Epidemic provides an upper bound on cost and a lower bound on delivery delay. Therefore, a deep understanding of these features is useful to make informed forwarding decisions. In this paper, we try to explore these features by observing the roles of different social relationships in the optimal paths through a set of real datasets. These datasets provide evidence that strangers have two sides in data forwarding process, and that the importance of strangers shows a decreasing trend along the forwarding paths. Using this heuristic knowledge, we propose STRON, a distributed and lightweight forwarding scheme. The distributed feature makes it very suitable for opportunistic scenarios and the low communication and computation features make it easy to be integrated with state-of-the-art work. The trace-driven simulations obviously confirm its effectiveness, especially in terms of packet delivery delay and cost. Routing is one of the challenging tasks in Delay Tolerant Networks (DTNs), due to the lack of global knowledge and sporadic contacts between nodes. Most existing studies take a greedy scheme in data forwarding process, ie, only nodes with higher utility values ​​than current They lack an in-depth investigation on the main features of the optimal paths in Epidemic. These features are vital to any forwarding scheme that tends to make a trade-off between packet delivery delay and cost. This is Since because Epidemic provides an upper bound on cost and a lower bound on delivery delay. Therefore, a deep understanding of these features is useful to make informed forwarding decisions. In this paper, we try to explore these features by observing the roles of different social relationships in the optimal paths through a set of real datasets. These datasets provide evidence that strangers have two sides in data forwarding process, and that the importance of s Using this heuristic knowledge, we propose STRON, a distributed and lightweight forwarding scheme. The distributed feature makes it very suitable for opportunistic scenarios and the low communication and computation features make it easy to be integrated with state-of-the-art work. The trace-driven simulations obviously confirm its effectiveness, especially in terms of packet delivery delay and cost.