Content area
Full text
Mathematical optimization is broadly applicable to the problems of minimizing costs, maximizing profits, and scheduling projects subject to time and financial constraints. Computer software implementing the techniques of mathematical optimization has been much in demand, and IBM has been a leader in providing such software. For many years, IBM's Mathematical Programming System Extended/370 (MPSX/370)(1) program product has been a standard tool used by optimization specialists for applications in a wide variety of business and government areas. However, the optimization field has grown in size, variety, and complexity of applications. New methods are needed to solve problems accurately and quickly.
The Optimization Center in the Mathematical Sciences Department of IBM Research has been continuously active in developing needed techniques, and there has been an ongoing partnership between the Mathematical Sciences Department and the Application Technology Center of the IBM Kingston High Performance Computing Solutions Development organization devoted to integrating Research's contributions into IBM engineeing and scientific products. This partnership had already produced the IBM Engineering and Scientific Subroutine Library (ESSL). Thus, when it became clear that a new kind of product was needed to support use of IBM systems for optimization applications, it was natural to turn to this partnership, and the Optimization Subroutine Library (OSL) is the result.
OSL was developed to provide practitioners with a suite of tools for manipulating the mathematical models and solving the resulting minimization and maximization problems of mathematical optimization. OSL is a departure from the optimization software products of the past. It has algorithmic breadth: having linear programming via both simplex and interior point approaches, mixed-integer programming, quadratic programming, and network problems. It has platform breadth: being available on most IBM workstation and mainframe systems with most of the operating system environments, including a low-end version running under the popular disk operating system from Microsoft Corporation. It is extensible, and it is not subject to problem size limitations other than those imposed by the platforms. Finally, it is readily adapted to provide solutions for vendors' products.
In this issue of the IBM Systems Journal the reader is introduced to the depth and breadth of OSL. In the remainder of this essay, we identify the problems that OSL addresses and give a moderately technical overview of the content of...