## FATCOP: A fault tolerant Condor-PVM mixed integer program solver. Mathematical Programming (1999)

Citations: | 27 - 4 self |

### BibTeX

@TECHREPORT{Chen99fatcop:a,

author = {Qun Chen and Michael and C. Ferris},

title = {FATCOP: A fault tolerant Condor-PVM mixed integer program solver. Mathematical Programming},

institution = {},

year = {1999}

}

### Years of Citing Articles

### OpenURL

### Abstract

Abstract. We describe FATCOP, a new parallel mixed integer program solver written in PVM. The implementation uses the Condor resource management system to provide a virtual machine composed of otherwise idle computers. The solver differs from previous parallel branch-and-bound codes by implementing a general purpose parallel mixed integer programming algorithm in an opportunistic multiple processor environment, as opposed to a conventional dedicated environment. It shows how to make effective use of resources as they become available while ensuring the program tolerates resource retreat. The solver performs well on test problems arising from real applications and is particularly useful for solving long running hard mixed integer programming problems.

### Citations

1057 | Condor - a hunter of idle workstations - Litzkow, Livny, et al. - 1988 |

1013 |
Using MPI: Portable Parallel Programming with the Message-Passing Interface
- Gropp, Lusk, et al.
- 1994
(Show Context)
Citation Context ...he system is the PVM library. It contains usercallable routines for message passing, process spawning, virtual machine modification and task coordination. A similar message passing environment is MPI =-=[14]-=-. Both systems center around a message-passing model, providing point-to-point as well as collective communication between distributed processes. The primary differences between the systems lie in the... |

630 | PVM: Parallel Virtual Machine A Users' Guide and Tutorial for Networked Parallel Computing - Geist - 1994 |

255 | A Users Guide
- Brooke, Kendrick, et al.
- 1998
(Show Context)
Citation Context ...ut as well as GAMS models. MPS [19] input format was originally introduced by IBM to express linear and integer programs in a standard way. All commercial LP and MIP software accept this format. GAMS =-=[4]-=- is a high-level modeling system for mathematical programming. It takes as input a description of a mathematical program in a form that people find reasonably natural and convenient, and allows the so... |

170 |
Condor-A hunter of idle workstations
- Litzkow, Livny, et al.
- 1988
(Show Context)
Citation Context ...urces in the machine will be “occupied” during the life cycle of the program. This is not a desirable situation when the machine is owned by a person different from the user of the MIP solver. Condor =-=[12, 20]-=- is a distributed resource management system that can help to overcome these problems. Condor manages large heterogeneous clusters of machines in an attempt to use the idle cycles of some users’ machi... |

123 |
GAMS: A User's Guide, The Scientific
- Brooke, Kendrick, et al.
- 1988
(Show Context)
Citation Context ...tunistic resources; • powerful MIP techniques including strong branching, pseudocost estimation searching, preprocessing, and cutting plane generation; • the ability to process both MPS [21] and GAMS =-=[7]-=- models; • the use of both CPLEX and SOPLEX as its LP solver. The remainder of this paper is organized as follows. Section 2 is a review of the standard MIP algorithm components of FATCOP that are imp... |

106 | A Worldwide Flock of Condors: Load Sharing amongWorkstation Clusters
- Epema, Livny, et al.
- 1996
(Show Context)
Citation Context ...urces in the machine will be “occupied” during the life cycle of the program. This is not a desirable situation when the machine is owned by a person different from the user of the MIP solver. Condor =-=[12, 20]-=- is a distributed resource management system that can help to overcome these problems. Condor manages large heterogeneous clusters of machines in an attempt to use the idle cycles of some users’ machi... |

90 | Analyzing Scalability of Parallel Algorithms and Architectures
- Kumar, Gupta
- 1991
(Show Context)
Citation Context ... program solver, written in PVM, that runs in the opportunistic computing environment provided by the Condor resource management system. Parallel BB algorithms for MIP have attracted many researchers =-=[9, 13, 15]-=-. Most parallel BB programs were developed to run using large centralized mainframes or big-iron supercomputers that are typically very expensive. Users of these facilities usually only have a certain... |

79 | An enabling framework for master-worker applications on the computational grid
- Goux, Kulkarni, et al.
- 1994
(Show Context)
Citation Context ...versity of Wisconsin, Madison, WI 53706 (ferris@ cs.wisc.edu). 10191020 QUN CHEN AND MICHAEL C. FERRIS the scheme first outlined in this work for other branch-and-bound applications, most notably in =-=[1, 16, 27]-=-. A parallel virtual machine (PVM) is a programming environment that allows a heterogeneous network of computers to appear as a single concurrent computational resource [14]. It provides a unified fra... |

76 |
et al., PVM: Parallel Virtual Machine A User's Guide and Tutorial for Networked Parallel Computing
- Geist
- 1994
(Show Context)
Citation Context ...es to solve large mixed integer programs. We believe that a parallel BB program developed to use this type of resource will become a widely used tool for such problems. PVM (parallel virtual machine) =-=[12]-=- is a parallel programming environment that allows a heterogeneous network of computers to appear as a single concurrent computational resource. It provides a unified framework within which parallel p... |

75 |
Parallel branch-and-bound algorithms: survey and synthesis
- Gendron, Crainic
- 1994
(Show Context)
Citation Context ... program solver, written in PVM, that runs in the opportunistic computing environment provided by the Condor resource management system. Parallel BB algorithms for MIP have attracted many researchers =-=[9, 13, 15]-=-. Most parallel BB programs were developed to run using large centralized mainframes or big-iron supercomputers that are typically very expensive. Users of these facilities usually only have a certain... |

63 | Optimized regional caching for on-demand data delivery
- Eager, Ferris, et al.
- 1999
(Show Context)
Citation Context ...urces Figure 5: Resource utilization for one run of FATCOP models, two problems formulated using GAMS are included in the test set. One problem vod3 is an application to video on demand system design =-=[8]-=-. The other problem prod1 is an application to product design [25]. These two problems are also insoluble in 10,000 nodes by the CPLEX default MIP solver. The sizes of the problem instances are shown ... |

51 | A computational study of search strategies for mixed integer programming
- Linderoth, Savelsbergh
- 1999
(Show Context)
Citation Context ...ninteresting [1]. Pseudocost attempts to estimate the change rate in objective function value associated with a particular branch. Since the pseudocost method is widely used and known to be efficient =-=[16]-=-, we set it as the default branching strategy and briefly describe the method here. We associate two quantities OE \Gamma j and OE + j with each integer variable x j that attempts to measure the per u... |

36 |
Parallel branch-and-bound algorithms for general mixed integer programming on the CM-5
- Eckstein
- 1994
(Show Context)
Citation Context ... program solver, written in PVM, that runs in the opportunistic computing environment provided by the Condor resource management system. Parallel BB algorithms for MIP have attracted many researchers =-=[9, 13, 15]-=-. Most parallel BB programs were developed to run using large centralized mainframes or big-iron supercomputers that are typically very expensive. Users of these facilities usually only have a certain... |

35 | Providing resource management services to parallel applications
- Pruyne, Livny
- 1994
(Show Context)
Citation Context ... passing environments are not able to use resources efficiently. Researchers at the University of Wisconsin-Madison recently developed a parallel programming framework which interfaces Condor and PVM =-=[22, 6]-=-. The reason to select PVM instead of MPI is that the implementation of MPI has no concept of process control, hence cannot handle resource addition and retreat in a opportunistic environment. Figure ... |

33 |
Preprocessing and probing for mixed integer programming problems
- Savelsbergh
- 1994
(Show Context)
Citation Context ... the size of a MIP problem as well as the integrality gap, i.e., the difference between the optimal solution value and its LP relaxation. More techniques for preprocessing and probing can be found in =-=[24]-=-. Without loss of generality we assume that the inequality currently under consideration is of the form: X j a j x jsb Define: l j The lower bound of variable x j ; u j The upper bound of variable x j... |

32 | Lifted cover inequalities for 0–1 integer programs: Computation - Gu, Nemhauser, et al. - 1998 |

27 |
Computer Solutions of Linear Programs
- Nazareth
- 1987
(Show Context)
Citation Context ... with a different but still integral value of x j would be bigger than Z , hence would necessarily be fathomed. 2.5 Interfaces FATCOP accepts industrial standard MPS input as well as GAMS models. MPS =-=[19]-=- input format was originally introduced by IBM to express linear and integer programs in a standard way. All commercial LP and MIP software accept this format. GAMS [4] is a high-level modeling system... |

26 | FATCOP 2.0: Advanced features in an opportunistic mixed integer programming solver
- Chen, Ferris, et al.
- 1999
(Show Context)
Citation Context ...low cover cuts and disjunctive cuts and investigating how much a worker processor should do before returning results to the master. Some of these issues have already been addressed in subsequent work =-=[8]-=-. Acknowledgments. The authors are grateful to Miron Livny and Michael Yoder for advice and assistance in using Condor and the Condor-PVM environment. The authors also wish to thank Jeff Linderoth for... |

20 | Cutting planes for integer programs with general integer variables
- Ceria, Cordier, et al.
- 1998
(Show Context)
Citation Context ...remained insoluble in a reasonable amount of time. Solving such problems may require some specialized techniques. For example recent work that combines BB with cutting planes has been very successful =-=[5]-=-. Adding cutting planes at the top of node for BB algorithm can substantially reduces the size of the search tree, and the overall computation time. This cut-and-branch framework might be used in a fu... |

19 |
Experiments in mixed-integer linear programming
- Bénichou, Gauthier, et al.
- 1971
(Show Context)
Citation Context ... but pushes the optimal value of the initial LP relaxation from 315.29 up to 3125.92. 2.2. Cutting planes and reduced cost fixing. It is well known that cutting planes can strengthen MIP formulations =-=[4]-=-. FATCOP generates knapsack cuts at each subproblem as described in [18]. There are about 10 models in MIPLIB for which knapsack cuts are useful. We again take p0548 as an example; the FATCOP code can... |

17 |
et al., “Condor - a hunter of idle workstations
- Litzkow
- 1988
(Show Context)
Citation Context ...esources in the machine will be "occupied" during the life cycle of the program. This is not a desirable situation when the machine is owned by a person different from the user of the progra=-=m. Condor [6, 10, 17]-=-, a distributed resource management system, can help to overcome the problems a normal PVM program runs into. Condor manages large heterogeneous clusters of machines in an attempt to make use of the i... |

16 | Parallel mixed integer programming
- Bixby, Cook, et al.
- 1995
(Show Context)
Citation Context .... Several reasonable criteria exist for selecting branching variables. FATCOP currently provides four variable selection options:1022 QUN CHEN AND MICHAEL C. FERRIS pseudocost [19], strong branching =-=[6]-=-, and maximum and minimum integer infeasibility [2]. Since the pseudocost method is widely used and known to be efficient, we set it as the default branching strategy. FATCOP can also accept user defi... |

13 |
Lifted Cover Inequalities for 0-1 Integer Programs
- Gu, Nemhauser, et al.
- 1998
(Show Context)
Citation Context ...up to 3125.92. 2.2. Cutting planes and reduced cost fixing. It is well known that cutting planes can strengthen MIP formulations [4]. FATCOP generates knapsack cuts at each subproblem as described in =-=[18]-=-. There are about 10 models in MIPLIB for which knapsack cuts are useful. We again take p0548 as an example; the FATCOP code can solve the model in 350 nodes with knapsack cuts applied at each node. H... |

12 |
Genetic Algorithms for Product Design
- Balakrishnan, Jacob
- 1996
(Show Context)
Citation Context ... It turns out that this information is critical to solve this problem in a reasonable amount of time [13]. For PROD problems, good integer feasible solutions were found using a genetic algorithm (GA) =-=[3]-=-1034 QUN CHEN AND MICHAEL C. FERRIS Table 4.5 Results for VOD and PROD problems. Name #rows #columns #nonzeros #integers time Eavg VOD1 107 306 1207 303 7.5 mins 11 VOD2 715 1513 7316 1510 5.2 mins 1... |

10 |
Branch-and-bound and parallel computation: A historical note
- Pruul, Nemhauser
- 1988
(Show Context)
Citation Context ...itten in PVM, that runs in the opportunistic computing environment provided by the Condor resource management system. Parallel branch-and-bound algorithms for MIP have attracted many researchers (see =-=[11, 15, 24]-=- and references therein). Most parallel branch-and-bound programs were developed for large centralized mainframes or supercomputers that are typically very expensive. Users of these facilities usually... |

5 |
et al.: A Worldwide Flock of Condors: Load Sharing among Workstation Clusters
- Epema
- 1996
(Show Context)
Citation Context ...esources in the machine will be "occupied" during the life cycle of the program. This is not a desirable situation when the machine is owned by a person different from the user of the progra=-=m. Condor [6, 10, 17]-=-, a distributed resource management system, can help to overcome the problems a normal PVM program runs into. Condor manages large heterogeneous clusters of machines in an attempt to make use of the i... |

4 |
2001): Solving optimization problems on computational grids. Optima 65
- Wright
(Show Context)
Citation Context ...versity of Wisconsin, Madison, WI 53706 (ferris@ cs.wisc.edu). 10191020 QUN CHEN AND MICHAEL C. FERRIS the scheme first outlined in this work for other branch-and-bound applications, most notably in =-=[1, 16, 27]-=-. A parallel virtual machine (PVM) is a programming environment that allows a heterogeneous network of computers to appear as a single concurrent computational resource [14]. It provides a unified fra... |

3 | a Mixed INTeger - MINTO - 1994 |

3 |
Solving Large Quadratic Assignment
- Anstreicher, Brixius, et al.
- 2000
(Show Context)
Citation Context ...versity of Wisconsin, Madison, WI 53706 (ferris@ cs.wisc.edu). 10191020 QUN CHEN AND MICHAEL C. FERRIS the scheme first outlined in this work for other branch-and-bound applications, most notably in =-=[1, 16, 27]-=-. A parallel virtual machine (PVM) is a programming environment that allows a heterogeneous network of computers to appear as a single concurrent computational resource [14]. It provides a unified fra... |

2 |
Mathematical Programming for Industrial Engineers
- Avriel, Golany
- 1996
(Show Context)
Citation Context ...x s.t. Axsb lsxsu x j 2 Z 8j 2 I where Z denotes the integers, A is an m \Theta n matrix, and I is a set of distinguished indices identifying the integer variables. Most integer programming textbooks =-=[1, 20]-=- describe the fundamental branch-and-bound algorithm for the above MIP problem. Basically, the method explores a binary tree of subproblems. The key steps of BB are summarized below and the algorithm ... |

2 | Report on the session on branch and bound/implicit enumeration, in discrete optimization - Balas, Martin - 1979 |

2 | et al. Practical solution of large scale mixed integer programming problems with UMPIRE - Forrest - 1979 |

2 |
An Optimization Framework for Product Design. submitted to Management Science
- Shi, Ólafsson, et al.
- 1998
(Show Context)
Citation Context ..., two problems formulated using GAMS are included in the test set. One problem vod3 is an application to video on demand system design [8]. The other problem prod1 is an application to product design =-=[25]-=-. These two problems are also insoluble in 10,000 nodes by the CPLEX default MIP solver. The sizes of the problem instances are shown in Table 1. 4.1 Resource Utilization In Wisconsin's Condor pool th... |

2 | Optimized Regional Caching for OnDemand Data - Eager, Ferris, et al. - 1999 |

2 | Models and solution for on-demand data delivery problems
- Ferris, Meyer
- 1999
(Show Context)
Citation Context ...t problems. FATCOP was used to solve two classes of problems arising from marketing and electronic data delivery. One class of problems, VOD, consists of applications to video-on-demand system design =-=[10, 13]-=-. The other class of problems, PROD, consists of applications to product design [26]. Two problems from each application were formulated as GAMS models. The size of the problem instances and results f... |

2 |
An Optimization Framework for Product Design. Management Science 47:1681–1692
- Shi, Olafsson, et al.
- 2001
(Show Context)
Citation Context ...electronic data delivery. One class of problems, VOD, consists of applications to video-on-demand system design [10, 13]. The other class of problems, PROD, consists of applications to product design =-=[26]-=-. Two problems from each application were formulated as GAMS models. The size of the problem instances and results found by FATCOP are reported in Table 4.5. Execution time is clock elapsed time and d... |

1 |
Documentation of the SOPLEX library. http://www.zib.de/Optimization/Software/Soplex
- Wunderling
(Show Context)
Citation Context ...rallel solver for mixed integer programs in Condor's opportunistic environment. It is implemented on top of SOPLEX , a simplex object-oriented linear programming solver developed by Roland Wunderling =-=[26]-=-. FATCOP is written in the C++ programming language with calls to PVM library. It is designed to make best use of participating resources managed by Condor while handling resource retreat carefully in... |

1 |
MathematicalProgramming for IndustrialEngineers
- Avriel, Golany
- 1996
(Show Context)
Citation Context ...ranching variables. FATCOP currently provides four variable selection options:1022 QUN CHEN AND MICHAEL C. FERRIS pseudocost [19], strong branching [6], and maximum and minimum integer infeasibility =-=[2]-=-. Since the pseudocost method is widely used and known to be efficient, we set it as the default branching strategy. FATCOP can also accept user defined priorities on integer variables. FATCOP provide... |

1 |
MIPLIB 3.0, available online at http://www.caam.rice.edu/˜bixby/miplib/miplib.html
- Bixby, Ceria, et al.
(Show Context)
Citation Context ...ssing may reduce the integrality gap, i.e., the difference between the optimal solution value and its LP relaxation as well as the size of an MIP problem. For example, for the model p0548 from MIPLIB =-=[5]-=-, an electronically available library of both pure and mixed integer programs arising from real applications, the FATCOP preprocessor can remove only 12 rows and 16 columns and modify 176 coefficients... |

1 |
available online at http://www.cplex.com
- Optimizer
(Show Context)
Citation Context ... general branch-and-bound codes for other applications. FATCOP is implemented on top of both SOPLEX, a public available simplex object-oriented linear programming solver [28], and the CPLEX LP solver =-=[9]-=-. FATCOP is written in the C++ programming language with calls to the PVM library. It is designed to make best use of participating resources managed by Condor while handling resource retreat carefull... |

1 |
Optimized regionalcaching for on-demand data delivery
- Eager, Ferris, et al.
- 1999
(Show Context)
Citation Context ...t problems. FATCOP was used to solve two classes of problems arising from marketing and electronic data delivery. One class of problems, VOD, consists of applications to video-on-demand system design =-=[10, 13]-=-. The other class of problems, PROD, consists of applications to product design [26]. Two problems from each application were formulated as GAMS models. The size of the problem instances and results f... |

1 |
A computationalstudy of search strategies for mixed integer programming
- Linderoth, Savelsbergh
- 1999
(Show Context)
Citation Context ...able and node selection. Several reasonable criteria exist for selecting branching variables. FATCOP currently provides four variable selection options:1022 QUN CHEN AND MICHAEL C. FERRIS pseudocost =-=[19]-=-, strong branching [6], and maximum and minimum integer infeasibility [2]. Since the pseudocost method is widely used and known to be efficient, we set it as the default branching strategy. FATCOP can... |

1 |
SOPLEX Library Documentation, available online at http://www.zib.de/Optimization/Software/Soplex
- Wunderling
(Show Context)
Citation Context ... can be incorporated into more general branch-and-bound codes for other applications. FATCOP is implemented on top of both SOPLEX, a public available simplex object-oriented linear programming solver =-=[28]-=-, and the CPLEX LP solver [9]. FATCOP is written in the C++ programming language with calls to the PVM library. It is designed to make best use of participating resources managed by Condor while handl... |