The invention discloses a high-efficient resource allocation method based on sensitivity in a reconfigurable network. The method comprises: computing a difference value between the maximum value and the secondary maximum value of the sensitivity Sen (i), comparing the difference value with a threshold value [alpha], and determining a candidate bottleneck node where a resource is required to be extra allocated according to a comparison result; computing a difference value between node CPU resource average utilization rate and link bandwidth resource average utilization rate, comparing the difference value with a threshold value [beta], and determining the type of the resource required to be extra allocated to the candidate bottleneck nods according to a comparison result; according to the determined candidate bottleneck node and the determined type of the resource required to be extra allocated, gradually increasing the corresponding recourse of the candidate bottleneck node by using an iterative method until a network average request acceptance rate converges or corresponding resource increment exceeds an upper limit. By effectively allocating extra resource to the candidate bottleneck node, the method solves a problem of a decrease in a request acceptance rate due to a resource bottleneck in a layered graph + capacity tracking algorithm.