The routing algorithm of SPF is widely distributed in the Internet. Since this routing algorithm is designed in order to improve throughput of each packet, it is not suitable for averaging load balance in the network. On the contrary, metabolic networks in the cell can realize load balance and achieve fault-tolerance by using enzymatic feedback mechanism. That is, a metabolic pathway in the cell is composed of a lot of enzymatic reaction steps in which biochemical reactant (substrate) is converted to the product by unique enzyme, and the product of a late step frequently acts as an inhibitor of the first committed step in this pathway (feedback control). This way, the end product of a pathway controls its own synthesis and prevents useless accumulation of intermediates and of end product. Recently, by mimicking enzymatic feedback mechanism in the cell, we have designed a fault-tolerant adaptive routing algorithm to avoid the partial and time-variant congestions in the network. We evaluated and compared the proposed algorithm with SPF and ECMP (Equal Cost Multi-path Protocol) by using the simulation of test data. The simulation results show that the proposed algorithm can remarkably improve both latency, load balance and fault tolerance. Since there are enormous numbers of nodes in the Internet, however, it is difficult to replace all existing nodes to the proposed nodes. In this paper, we shall propose an efficient method for the allocation of adaptive nodes in random and a scale-free network composed of 100 nodes. Examined the time-variant traffic at each node, and only focused on around 10% top ranked heavy-traffic nodes, we replace such nodes to our proposed adaptive nodes. By doing this, we could design a fault-tolerant adaptive routing, which can dynamically average load-balance within the network.