Originally posted by: gcak

Hi everyone. I am working on a hub cover problem for my thesis and trying to solve an mip. Actually, I am using GAMS for solution and GAMS calls CPLEX in order to get an optimal solution.
The problem has 25 candidate sites for opening hubs and in the optimal solution five hubs are opened. The optimal is obtained in 9781 seconds. In order to decrease the solution time, I add a constraint saying that at least three hubs should be opened. However, the solution time becomes 78270 seconds in this case. When i add a constraint saying that at most ten hubs should be opened, the solution time becomes 69760 seconds. The same optimal solution is obtained in all cases. My advisor says these should decrease the solution time. However, as seen, the solution time increses considereably. Does anybody have an idea about that?
By the way, I am new to here and I am not sure whether this is the appropriate place for this question. In addition, this is a CPLEX forum but I have not coded the program in CPLEX. I am calling CPLEX from within GAMS. So, I am sorry if this question is irrelevant for this place. I can send the GAMS version of the program if you wish. Thanks a lot.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

The additional constraints you add shrink the space of feasible solutions. This may or may not help CPLEX to find the optimal solution.
In general, you cannot say that adding these constraints makes the search faster. The "might" but that there is no guarantee that they "will".

For example, assume that without the additional constraints you find a feasible solution early in the search. This will allow CPLEX allow you to prune nodes, run improvement heuristics etc.
With the additional constraints finding a feasible solution may take longer time and CPLEX will may have to explore nodes that were pruned in the other case.
Does GAMMS produce any log files that show when infeasible solutions were found?

What might work better than adding a constraint 'hubs <= 10' is providing a solution start solution that uses only 10 hubs.
#CPLEXOptimizers
#DecisionOptimization