CPLEX 12.6 Announce
JeanFrancoisPuget 2700028FGP Visits (8916)
Update on June 28 2015. A more recent release of CPLEX is now available, namely CPLEX 12.6.2.
We are proud to announce release 12.6 of IBM ILOG CPLEX Optimization Studio. Planned release dates are
Version 12.6 delivers:
I cannot cover all of the new features int this post but here are some details. I will refer readers to the announcement letter for more information.
Parallel Distributed MIP
CPLEX 12.6 offers more support for the solution in parallel of mixed integer programs (MIPs) in a distributed computing environment. This new feature, known as distributed parallel MIP optimization, is a mode of running CPLEX that harnesses the power of multiple computers or of multiple nodes inside a single computer to achieve better performance on some MIP problems. This is a significant addition to the previous MIP solver that could only run on shared memory machines.
Distributed MIP is based on the CPLEX remote object for distributed parallel optimization introduced in CPLEX V12.5.1. Like the remote object, distributed MIP can potentially use any of the transport protocols available in the your computing environment. In order to specify to CPLEX which machines to use in addition to the master, you provide a file, written in XML and known as a Virtual Machine Configuration (VMC), to declare the transport protocol to use, the addresses of the
Distributed parallel MIP optimization operates in two phases: first, a ramp-up phase, in which each worker applies different parameter settings to the same problem as the other workers; then, in the remainder of the solve (the second phase), each worker works in one part of a common MIP tree. Each worker communicates what it finds to the (unique) master node, which acts as the conductor or coordinator for the whole process.
Distributed MIP can boost performance for some models compared to the shared memory MIP algorithm. However, Distributed MIP can also be of no help for some other models, i.e. it won't provide better performance than the shared memory MIP algorithm. We have some clues about which models are in the first category vs the second category but it is too early to provide general guidance. The only way to know for sure is to try it on your models! As explained above, moving from shared memory MIP algorithm to the distributed MIP algorithm is quite easy..
Solving nonconvex QP and MIQP
CPLEX V12.6 is capable of solving a quadratic program (a problem with linear constraints and one or more quadratic terms in the objective function) to global optimality. CPLEX finds this globally optimal solution regardless of whether or not the program is convex in the case of a minimization, or equivalently, whether or not the program is concave in the case of a maximization. In addition to the linear constraints and quadratic objective terms, the problem can also include integer constraints, and CPLEX V12.6 can solve the MIQP to global optimality.
In terms of its matrix, such a problem is also sometimes known as an indefinite MIQP.
Previous CPLEX versions could solve definite positive QPs to optimality. For indefinite QPs the returned solution was either a local optimum or a solution satisfying KKT conditions, but optimality was not guaranteed.
With the new code assist features of the CPLEX Optimization Studio IDE, users will be able to speed up the development of their optimization models. The IDE is now providing "auto" completion for functions and structures in OPL models, contextual search among OPL projects, code formatting and other useful features to ease the development and debugging of optimization models based on the OPL language.
Several enhancements are available in this release