An empirical evaluation of the KORBX’ª( àalgorithms for military airlift applications

KORBX’ª( (a registered trademark of AT&T) is AT&T’s new system for solving large-scale linear programs. The system consists of both hardware, which uses parallel processing technology configured with 256 MB of memory, and software which exploits the design and resources of this modern hardware. The KORBX linear programming software system contains four algorithms which are variations of the interior point method of Narendra Karmarkar. The primal, dual, primal-dual, and power series algorithms were empirically evaluated on a set of linear programming application models being used by the staff of the Military Airlift Command at Scott Air Force Base. For calibration purposes, a set of smaller test problems were also run using MPSX and XMP; and some pure network problems were solved using NETFLO, MPSX, and XMP.