In Dense Wavelength Division Multiplexing (DWDM) technologies, the optimal packet scheduling issue is commonly encountered in multiple channel networks. NP-hard problems deal with finding a way to rearrange packets from multiple channels into a finite and rare channel. Genetic Algorithm (GA) is one of the most efficient ways to solve these issues. We hope to find a better solution to our task through the GA characteristics of multiprocessor searching and survival of the fittest. Therefore, a modified and achievable hardware architecture of GA is presented in this paper. This architecture can increase both the speed of packet scheduling and the efficiency of DWDM in Optical Communication Networks.