Resumen
We consider algorithms for solving integer optimization problems with quasi--block structure. Modern decomposition approaches are analyzed. We study Finkelstein decomposition and it variations for discovering quasi--block structure. Efficiency of local elimination algorithm for large-scale problem is analyzed. Specific details of application of parametric optimization are provided. Dependency of order of solving subproblems on the algorithm performance is studied. We provide results of numerical experiments for solving large--scale linear programming problems by exact, approximate, or heuristic algorithms. Also we present experiments for parallelization of local elimination algorithm using grid computing approach. We discuss some problems which cannot be solved efficiently without parallelization techniques.