A joint capacity allocation approach is proposed to optimize spare capacity in multi-layer survivable networks. Depending on a full overview of all network layers, the joint capacity allocation approach can realize the maximum sharing of spare capacity among layers. The spare capacity allocation problem in multi-layer survivable networks is formulated through Integer Linear Programming, while for large-scale networks, a genetic algorithm is also proposed for possible optimal solution. Numerical results illustrate that the proposed approach is much more cost-effective. Due to the introduction of coordination among different layers, the proposed approach can also avoid the backhauling problem.