Job shop is one of the most popular manufacturing environments in practice and is extensively discussed in literatures. Most of the literatures focus on one machine setup at each shop, which usually does not happen in practice for the purpose of workload balancing and make-span shortening. Therefore, this study aims to minimize total earliness and tardiness costs of a multi-processor job shop scheduling problem with due windows. Ant colony optimization (ACO) is deployed to solve the problem. Simulation data testing results show that ACO has similar solutions to integer programming in small scale problems, and robust, effective and time-efficient solutions in large scale problems.