Generally speaking, in the practical optical network, evaluating network performance is used by multiple objective functions. But there is no one which can make the multiple objective functions to achieve the optimization at the same time. To solve this problem, this paper proposes a routing optimization algorithm based on co-evolution and chaos genetic algorithm, which takes the network time delay, the average risk and the risk balance as the objective functions. The algorithm not only ensures the services QoS requirements, but also reduces the network average risk, increases the risk balance degree, and improves the ability to resist the network attack by adjusting the routing of service which carried on high risk link.