Benchmark Instances in .mps format

MPS is one of the standard formats for expressing mixed integer linear programs. This site aims to provide some of the well-known OR-benchmark instances in this format. All of these instances were collected from publicly accessible domains and links to the original sources are provided along with.

The first two columns of the following table give information about the problem domain of the instances. The third column gives the number of instances in the collection which are available in the .mps format. The “ASCII” column contains links to the instances in text format (if available), where as the “MPS” column contains links to the instances in .mps format. An entry of “email” in the “MPS” column indicates that the .mps instances are accessible only via email to anureet “at” cmu “dot” edu. The ASCII-to-mps format conversion code is available for some of the instances as indicated in the column titled “Code”. The format conversion code uses COIN-OR libraries and is otherwise self-contained. In order to assess the computational hardness of these instances, the result of running CPLEX for a fixed period of time is also provided in the form of two files – input_configuration.txt and instance_info.txt. The former contains the CPLEX configuration which was used (time limit and  MIP emphasis) and the later contains a summary of the experiment for each instance which includes the number of (rows, columns, binary variables, integer variables), best-solution value & best-bound found by CPLEX, number of branch and bounds enumerated and the version of CPLEX used for the experiment. This information can be accessed via the “CPX info” column of the table.

Links suffixed with * point to data/code developed by Anureet Saxena, which is made available for research purposes.  The author, however, bears no responsibility to any kind of loss incurred due to direct or indirect use of data provided on this site.  If you have benchmark instances in .mps format and would like to post them on this site, please write to anureet “at” cmu “dot” edu.

ORLIB
Type Sub-Type # Instances ASCII MPS Code CPX Info
Capacitated Warehouse Location Set 1 37 www www * NA www *
  Set 2 12 www email* NA www *
Capacitated p-Median   20 www www * www * www *
AirLanding Scheduling    25 www www * www * www *
Set Covering   80 www www * NA www *
BCOL
Type Sub-Type # Instances ASCII MPS Code CPX Info
Mixed Knapsack  Unbounded 30 NA www NA NA
  Bounded 90 NA www NA NA
Fixed Charge Network Flow    20 NA www NA www*
Capacitated Lot Sizing   100 NA www NA www*
Capacitated Network Design Arc-Set   35 NA www NA www*
Network Flow with Additive Variable Upper Bounds   60 NA www NA www*
Miscellaneous Instances
Type Sub-Type # Instances ASCII MPS Code CPX Info
Holmberg Capacitated SSCFLP    71 www www * www * www*
Generalized Assignment Problems   57 www www * NA www*
Lot-Sizing with Backlogging (Kucukyavuz and Pochet)   65 NA www  NA www*
Fixed Charge Transportation Problems   32 NA www  NA NA
SNDlib (selected problems by C. Raack)   54 NA www  NA www*
Network Design Instances (contributed by P. Belotti) Small-Medium 35 NA www  NA www *
Related Links
Name Comment
MIPLIB 2003 Well-Known Library of general mixed integer linear programs
MIPLIB 3.0 Predecessor of MIPLIB 2003
CORAL MIP Instances General mixed integer linear programs
Decision Tree for Optimization Software Links to several structured/unstructured (non)-linear benchmark instances
Optimization and Operations Research Useful Resources
TSPLIB Well-Known Library of TSP Instances
QAPLIB Well-Known Library of Quadratic Assignment Problem Instances
LOTSIZELIB Lot Sizing Problem: A Library of Models and Matrices
MultiCommodity Flow Problems Library of several versions of multi-commodity flow problems 
SYMPHONY Open Source code for Branch&Cut, Benchmark instances for several routing problems
Benchmarks with Hidden Optimal Solutions Pseudo-Boolean (0-1 Integer Programming) Benchmarks with Hidden Optimum Solutions
SNDlib Library of test instances for Survivable fixed telecommunication Network Design.