Aiming at the frequently-occurring sudden-onset disasters and insufficient emergency traffic guarantee ability in the world, this paper analyzes the influences of sudden-onset disasters on the urban traffic network and travel behaviors, and proposes a timing optimization model of traffic signals for emergency evacuation in sudden-onset disaster areas. The model, which is based on the dynamic traffic assignment theory and the route choice beha-viors of drivers in sudden-onset disasters, is then optimized by using a parallel ant colony algorithm to accelerate the speed of solving the problem and improve the global convergence ability. Finally, a cluster computing platform is constructed to test the proposed model and algorithm. The results show that the parallel ant colony algorithm helps to obtain better time performance and higher optimization efficiency, thus meeting the real-time demands of on-line decision support for the emergency traffic guarantee system.