Principal component analysis was exploited to extract the principal features of endtoend network traffic and to attain the initial estiamtion results. This results are taken as the prior value of genetic algorithm to overcome its sensitiveness to the prior value. The estiamtion biases of link traffic is regarded as the fitness function of genetic algorithm. The crossover and mutation probability functions are built to control its corossover and mutation processes. The appropriate iterative funtion with contraints is built. The genetic algorithm is used to attain the endtoend traffic estimation results in the iterative way. Simulation results show that the proposed method is feasible.