Can SAS/OR create an MPS text file?

[cuencomi]

I ran out of memory after 2 hours with 24% relative gap.

NOTE: Problem generation will use 4 threads.
NOTE: The problem has 162671 variables (0 free, 0 fixed).
NOTE: The problem uses 2 implicit variables.
NOTE: The problem has 162671 binary and 0 integer variables.
NOTE: The problem has 2931 linear constraints (113 LE, 2818 EQ, 0 GE, 0 range).
NOTE: The problem has 325229 linear constraint coefficients.
NOTE: The problem has 0 nonlinear constraints (0 LE, 0 EQ, 0 GE, 0 range).
NOTE: The remaining solution time after problem generation and solver initialization is 12590.68 seconds.
NOTE: The MILP presolver value AUTOMATIC is applied.
NOTE: The MILP presolver removed 2093 variables and 8 constraints.
NOTE: The MILP presolver removed 4185 constraint coefficients.
NOTE: The MILP presolver modified 16 constraint coefficients.
NOTE: The presolved problem has 160578 variables, 2923 constraints, and 321044 constraint coefficients.
NOTE: The MILP solver is called.
NOTE: The parallel Branch and Cut algorithm is used.
NOTE: The Branch and Cut algorithm is using up to 4 threads.
Node Active Sols BestInteger BestBound Gap Time
0 1 3 1810300 2950.4854111 61256.0% 47
0 1 3 1810300 1050302 72.36% 48
0 1 3 1810300 1062232 70.42% 75
0 1 3 1810300 1066621 69.72% 96
0 1 3 1810300 1068548 69.42% 124
0 1 3 1810300 1069763 69.22% 162
0 1 3 1810300 1070563 69.10% 211
0 1 3 1810300 1071124 69.01% 276
0 1 3 1810300 1071673 68.92% 357
0 1 3 1810300 1071984 68.87% 452
0 1 3 1810300 1072216 68.84% 591
0 1 3 1810300 1072392 68.81% 691
0 1 3 1810300 1072530 68.79% 800
0 1 3 1810300 1072682 68.76% 905
0 1 3 1810300 1072829 68.74% 1014
NOTE: The MILP solver added 1988 cuts with 33934 cut coefficients at the root.
1 2 3 1810300 1072829 68.74% 1030

751 1 3 1810300 1073363 68.66% 1699
2842 2077 3 1810300 1073575 68.62% 2000
8112 7337 3 1810300 1073988 68.56% 2500
12671 11890 3 1810300 1074031 68.55% 3000
15756 14970 4 1497585 1074139 39.42% 3232
18259 17468 4 1497585 1074357 39.39% 3500
23711 22912 4 1497585 1074442 39.38% 4002
28663 27857 4 1497585 1074500 39.38% 4500
32773 31960 5 1451508 1074638 35.07% 4905
33029 32214 5 1451508 1074677 35.06% 5002
38138 37313 5 1451508 1074710 35.06% 5500
43450 42619 5 1451508 1074724 35.06% 6000
49459 48619 5 1451508 1074746 35.06% 6500
55063 54213 5 1451508 1074755 35.05% 7000
60147 59290 6 1421391 1074771 32.25% 7425
60674 59814 7 1328658 1074780 23.62% 7492
60971 60111 7 1328658 1074780 23.62% 7500
ERROR: Out of memory.
NOTE: Objective of the best integer solution found = 1328657.5189.

NOTE: Total profiled time is 7851.03 seconds.
NOTE: PROCEDURE OPTMODEL used (Total process time):
real time 2:10:49.75
user cpu time 6:30:06.98
system cpu time 2:04.75
memory 8129798.53k
OS Memory 8187904.00k
Timestamp 04/14/2019 02:19:53 PM
Step Count 22 Switch Count 12365
Page Faults 971
Page Reclaims 0
Page Swaps 0
Voluntary Context Switches 817916
Involuntary Context Switches 625004
Block Input Operations 1
Block Output Operations 102

[RobPratt]

Three suggestions:

1. Use the SAS MEMSIZE option to increase the amount of memory available.
2. Upgrade to SAS/OR 15.1.
3. Use the NODESEL=DEPTH MILP solver option. This approach will use less memory but will likely take longer.

[cuencomi]

Increasing the memory solved the "out of memory" problem. However, after 5 hours the solver could only get down to 20% gap while yours got down to 1% in 2.5 hours. Do you think the difference can be attributed to using 15.1 version?  That will have to wait until we upgrade to that version. 

103454 102545 9 1314332 1075017 22.26% 15504
107197 106280 9 1314332 1075034 22.26% 16001
110686 109762 10 1305394 1075042 21.43% 16417
111106 110182 10 1305394 1075042 21.43% 16503
114870 113942 11 1296700 1075044 20.62% 16853
116177 115247 11 1296700 1075050 20.62% 17001
121863 120921 11 1296700 1075051 20.62% 17500
125398 124447 11 1296700 1075071 20.62% 17989
NOTE: Real time limit reached.
NOTE: Objective of the best integer solution found = 1296700.2137.

NOTE: PROCEDURE OPTMODEL used (Total process time):
real time 5:00:48.50
user cpu time 14:30:25.59
system cpu time 4:11.71
memory 13788369.93k
OS Memory 13848088.00k
Timestamp 04/16/2019 01:43:57 AM
Step Count 30 Switch Count 28432
Page Faults 628
Page Reclaims 0
Page Swaps 0
Voluntary Context Switches 1867507
Involuntary Context Switches 1369205
Block Input Operations 0
Block Output Operations 100

[RobPratt]

By using the following SOLVE statement with SAS/OR 14.3, I got to 1% gap in 4 hours:

   solve with milp / relobjgap=0.01 heuristics=aggressive;

506845 289747 99 1086819 1076058 1.00% 14670
506937 289839 99 1086819 1076058 1.00% 14675
507053 289955 99 1086819 1076058 1.00% 14680
507184 290086 99 1086819 1076058 1.00% 14685
507309 290211 99 1086819 1076058 1.00% 14690
507426 290328 99 1086819 1076058 1.00% 14695
507552 290454 99 1086819 1076058 1.00% 14700
507655 290557 99 1086819 1076058 1.00% 14705
507764 290666 99 1086819 1076058 1.00% 14710
507776 290677 99 1086819 1076058 1.00% 14710
NOTE: Optimal within relative gap.
NOTE: Objective = 1086818.8613.
...
NOTE: PROCEDURE OPTMODEL used (Total process time):
real time 4:06:05.18
user cpu time 11:20:27.75
system cpu time 2:14.91
memory 29682539.90k
OS Memory 29721888.00k
Timestamp 04/17/2019 03:11:19 PM
Step Count 1 Switch Count 5998

Also, the gap was below 10% within the first 50 minutes:

123500 122457 16 1182907 1074890 10.05% 2960
123547 122500 17 1177158 1074890 9.51% 2962

[cuencomi]

Since you were successful with using SAS/OR 14.3 also, the issue may be with the computer hardware? I am using SAS/OR on a Unix server. I will check with our hardware folks why I can't get below 20% gap in reasonable time. What do you think?