Compared with the Genetic Algorithm, a multi-population genetic algorithm has an enhancement in performance, but for a job shop scheduling problem, which has a lot of local optima, it also has the shortcomings of an easy-to-fall into local optima and a poor ability of local search. Therefore, an effective algorithm is proposed to solve job shop scheduling problem. The proposed algorithm, based on multi-population genetic algorithm, involves the strategy of memory-base and a mechanism of the Lamarckian evolution. Not only does the memory-base make individuals between sub-populations exchange information, but it can maintain the diversity of the population. The local search operator, based on Lamarckian evolution, is adupted to enhance the individual’s ability of local search. The simulated annealing algorithm that has a stronger ability to jump out local optima than the genetic algorithm is used, thus, alleviated the problem and enhances the performance of the algorithm for job shop scheduling. The experimental results on the well-known benchmark instances show the proposed algorithm is very effective in solving job shop scheduling problems.