All three of the optimizers offered by Haverly Systems run as fully-integrated components of the GRTMPS and H/Sched systems; their operation is seamless and transparent to the planner and scheduler. When H/Xpress or H/Cplex are combined with modern computer hardware, very large LP models (>40,000 rows), which recently required the computing power of a workstation or mainframe, can now be solved on a desktop PC. Each of the optimizers described below are widely used around the world.

H/Xpress is Haverly's version of the high-performance optimizer developed by Dash Optimization. The standard version of H/Xpress features all of the latest technology including MIP, SOS and Barrier/Interior Point.

H/Cplex is Haverly's version of the high-performance optimizer developed by Cplex/ILOG. This optimizer is also available with MIP, SOS and Barrier/Interior Point.

HS/LP is a robust, reliable optimizer developed by Haverly Systems which is suitable for smaller models. For more information about HS/LP, click here.

	Simplex solver	Barrier solver	Mixed Integer Programm'g	Special Ordered Sets
H/Xpress	yes, high- performance	yes, standard	yes, standard	yes, standard
H/Cplex	yes, high- performance	yes, extra cost	yes, extra cost	included with MIP option
HS/LP	yes, limited to 8,128 rows	no	yes, standard	no

Simplex is the original algorithm developed by Dr. George Dantzig to solve Linear Programming problems. Optimizers like H/Xpress and H/Cplex offer very high performance by utilizing the newest Simplex technology. It is common for H/Xpress and H/Cplex to solve a medium-size model four or five times faster than HS/LP using Simplex.

Barrier/Interior Point is an alternate algorithm used to solve larger models. Performance is highly dependent on whether your particular model is numerically compatible with the Interior algorithm. When it is compatible, Barrier/Interior Point can outperform high-performance Simplex by a factor of two, three or more.

Mixed Integer Programming (MIP) is an LP technique that ensures selected LP variables solve to integer values. For example, a model can be restricted to purchasing whole tanker loads of crude.

A Special Ordered Set (SOS) is a special case of the Mixed Integer Programming LP technique. SOS is used to limit the number of active LP vectors among a set of two or more LP vectors in an optimal solution. For example, if a reformer unit is modeled with four modes of operation (severity = 94, 96, 98 and 100), SOS can be used to restrict the reformer operation to two adjacent operational modes (e.g., severities 96 and 98). This is a powerful modeling technique yet is not difficult to implement in an application like GRTMPS.

Related topics:
GRTMPS LP System
OMNI Modeling Language

Return to the top of page.

For more information about Haverly Systems, click here.