In this paper, an optimal routing strategy called TM3NORA for the Internet of Things’ Machine-to-Machine network is proposed. In the Internet of Things, devices or nodes are clustered to create a Machine-to-Machine network, which can be considered as a small-world network. The network model concerned is a Time-constrained Multi-packet Multi-state Machine-to-machine network (TM3N). A TM3N has four characteristics: (1) each arc employs two attributes, capacity and lead time; (2) the arc capacity is stochastic; (3) different packets consume the arc capacity differently; (4) the delivery has to be completed within the promised time window. Essentially, the TM3NORA generates a small-world route map, reduces the route map, obtains the optimal route set that can be used to deliver all types of packets from the source node to the sink node within the desired time window, and evaluates the small-world route map’s overall reliability that all types of packets can arrive at the sink node within the time constraint. The efficiency and accuracy of the TM3NORA are validated through a numerical example of a six-node small-world route map.