In Dense Wavelength Division Multiplexing (DWDM) technologies, the optimal packet scheduling is a common encounter issue in multiple channels network. NP-hard problem deals with finding a way to rearrange packets in multiple channels into a finite and rare channel. Genetic algorithm (GA) is one of the most efficient ways to solve this issue. We hope to find a better solution to our task through the GA characteristics of multiprocessor searching and survivor of the fittest. Therefore, a modified and achievable hardware architecture of GA is presented in this paper. This architecture can increase the schedule speed of packet scheduling also can promote the efficiency of DWDM in Optical Communication Networks.