The importance of Master Production Schedule (MPS) and Group Technology (GT) is increasing. A master production scheduling system is developed for a GT cell. Aggregate production planning subsystem handles the problem where there are conflicting goals from different functions. A goal programming model is used to solve this problem under a GT environment and an illustrative example is discussed;A master production scheduling subsystem handles the multi-item lot sizing problem where overloading or subcontracting is allowed. Three heuristics are discussed to find a near optimal solution. Method A is the traditional period-by-period method and Method B is the item-by-item method where the shortest path algorithm is applied. These two methods generate a Wagner-Whitin (W-W) type schedule and are search methods for the global optima. Method C is a tree search method which gives non W-W type schedule and searches local optima from a good W-W type schedule which can be derived from Method A or Method B;A diverse set of test problems is defined and real data from industry along with systematically determined data are used to verify these three methods. The solution standard of these three methods is a left threshold parameter of an unknown distribution;If the time horizon is H periods, it takes O(H) time for method A and O(H('3)) time for Method B. Even though the search time of Method B is O(H('3)), Method B gives more reliable solutions than the traditional Method A for small size problems. Method A is better than Method B for medium and large size problems. A Wilcoxon's signed-rank test shows that there is strong evidence that the setup cost from Method A is less than that from Method B and that the holding cost from Method A is larger than that from Method B. There is no significant difference between Method A and Method B for overload cost;Even though an aggregate production planning model and three methods for master production scheduling are proposed for a GT environment, the aggregate production planning model can be easily revised for other environments and the three methods can be applied to other situations where the production lead time is short.