Originally posted by: JerryGuo

I'm solving a MIP. The root node relaxation only costed about 12 seconds to solve, as shown in the attached screen shot. However, after that, we just stayed at the root node for around 4 mins, and then it suddenly said that the problem was solved to optimality (actually, it should be a feasible integer solution, as the objective here is minimizing 0).

Since I switched off the heuristics (HeurFreq -1 and FPHeur -1), I don't think the solver was trying different heuristic algorithms to find a feasible integer solution during that 4 mins. And you can see the attached screen shot that there are no cuts generated during the whole solving process. So, I'm really confused what was the solver doing during that a long period of time? Moreover, since no heuristics, no cuts and no branchings were applied, how did the solver finally figure out the feasible solution? 

Are you sure the additional 4 minutes were spent at the root and not before in presolve? Can you show the full log?

Originally posted by: JerryGuo

Thanks for the reply, Daniel!

Yes, I'm pretty sure that the additional 4 mins were not taken by the presolve.

Attached is the full log, and you can see that the solver first pre-processed (pre-solve) the model and reduced the size a lot. Then it just solved the LP relaxation of the root node (by dual simplex, I think).

In total, these two steps costed only a few seconds, but after that it just got stuck at the first row of the nodes' log for about 4 mins.

It seems that the solver was doing a great number of computations during that 4 mins, but as you can see, the count of the iterations did not increase (still 10) when the feasible integer solution was figured out.

So it looks really weird to me for understanding what was the solver doing and how did it get that feasible solution at the end.

OK, that indeed looks odd. Can you share the model and non-default parameter settings you are using? If you don't want to share here, can you share with daniel(dot)junglas(at)de(dot)ibm(dot)com? We can then figure out what CPLEX is doing.

Originally posted by: JerryGuo

Hi Daniel,

I just sent you an email with the model (a .lp file) of the instance we're solving and the non-default parameter settings (a .txt file) attached.

Please let me know, if you need additional info.

Thank you so much!