The network function virtualization (NFV) paradigm focuses on increasing manageability and scalability of modern complex heterogeneous networks and network services by decoupling the network functions and hosting devices. However, as new promising solutions become available, the need for availability and reliability techniques grow, particularly for large-scale and interdependent scenarios. Therefore this study proposes a meta-heuristic genetic algorithm scheme to deploy “risk-aware” virtual function mapping and traffic routing to improve the reliability of user services as well as reduce deployment and routing costs. Furthermore, this solution is compared with two other “risk-aware” survivable schemes in order to evaluate its accuracy, i.e., integer linear programming (ILP) and greedy heuristic solutions.