A Grid-enabled Branch and Bound Algorithm for Solving Challenging Combinatorial Optimization Problems Authors: M. Mezmaz, N. Melab and E-G. Talbi Presented.

Slides:



Advertisements
Similar presentations
Optimal Bus Sequencing for Escape Routing in Dense PCBs H.Kong, T.Yan, M.D.F.Wong and M.M.Ozdal Department of ECE, University of Illinois at U-C ICCAD.
Advertisements

Social network partition Presenter: Xiaofei Cao Partick Berg.
Resource Management §A resource can be a logical, such as a shared file, or physical, such as a CPU (a node of the distributed system). One of the functions.
C4.5 algorithm Let the classes be denoted {C1, C2,…, Ck}. There are three possibilities for the content of the set of training samples T in the given node.
Lecture 3: Parallel Algorithm Design
CS 484. Discrete Optimization Problems A discrete optimization problem can be expressed as (S, f) S is the set of all feasible solutions f is the cost.
Clustering and Load Balancing Optimization for Redundant Content Removal Shanzhong Zhu (Ask.com) Alexandra Potapova, Maha Alabduljalil (Univ. of California.
The Theory of NP-Completeness
Progress in Linear Programming Based Branch-and-Bound Algorithms
Branch & Bound Algorithms
Gossip Algorithms and Implementing a Cluster/Grid Information service MsSys Course Amar Lior and Barak Amnon.
© The McGraw-Hill Companies, Inc., Chapter 8 The Theory of NP-Completeness.
Algorithm Strategies Nelson Padua-Perez Chau-Wen Tseng Department of Computer Science University of Maryland, College Park.
Distributed Computations
Fault-tolerant Adaptive Divisible Load Scheduling Xuan Lin, Sumanth J. V. Acknowledge: a few slides of DLT are from Thomas Robertazzi ’ s presentation.
SPIE Vision Geometry - July '99 Even faster point set pattern matching in 3-d Niagara University and SUNY - Buffalo Laurence Boxer Research.
EA* A Hybrid Approach Robbie Hanson. What is it?  The A* algorithm, using an EA for the heuristic.  An efficient way of partitioning the search space.
Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy under contract.
High Performance Computing 1 Parallelization Strategies and Load Balancing Some material borrowed from lectures of J. Demmel, UC Berkeley.
1 Branch and Bound Searching Strategies 2 Branch-and-bound strategy 2 mechanisms: A mechanism to generate branches A mechanism to generate a bound so.
Ch 13 – Backtracking + Branch-and-Bound
CS 584. Discrete Optimization Problems A discrete optimization problem can be expressed as (S, f) S is the set of all feasible solutions f is the cost.
1 Tree Searching Strategies. 2 The procedure of solving many problems may be represented by trees. Therefore the solving of these problems becomes a tree.
Branch and Bound Algorithm for Solving Integer Linear Programming
Ant Colony Optimization Algorithms for the Traveling Salesman Problem ACO Kristie Simpson EE536: Advanced Artificial Intelligence Montana State.
Distributed Computations MapReduce
Defining Polynomials p 1 (n) is the bound on the length of an input pair p 2 (n) is the bound on the running time of f p 3 (n) is a bound on the number.
Advanced Topics: MapReduce ECE 454 Computer Systems Programming Topics: Reductions Implemented in Distributed Frameworks Distributed Key-Value Stores Hadoop.
MGR: Multi-Level Global Router Yue Xu and Chris Chu Department of Electrical and Computer Engineering Iowa State University ICCAD
Pregel: A System for Large-Scale Graph Processing
By: Jeffrey Dean & Sanjay Ghemawat Presented by: Warunika Ranaweera Supervised by: Dr. Nalin Ranasinghe.
CS 221 – May 13 Review chapter 1 Lab – Show me your C programs – Black spaghetti – connect remaining machines – Be able to ping, ssh, and transfer files.
Combinatorial Optimization on the Computational Grid Experiments on Grid5000 Nouredine Melab ( ) Member of Grid5000 steering committee Laboratoire.
林俊宏 Parallel Association Rule Mining based on FI-Growth Algorithm Bundit Manaskasemsak, Nunnapus Benjamas, Arnon Rungsawang.
1 The Theory of NP-Completeness 2012/11/6 P: the class of problems which can be solved by a deterministic polynomial algorithm. NP : the class of decision.
1 The Map-Reduce Framework Compiled by Mark Silberstein, using slides from Dan Weld’s class at U. Washington, Yaniv Carmeli and some other.
CSCE350 Algorithms and Data Structure Lecture 17 Jianjun Hu Department of Computer Science and Engineering University of South Carolina
MapReduce M/R slides adapted from those of Jeff Dean’s.
CS 5204 (FALL 2005)1 Leases: An Efficient Fault Tolerant Mechanism for Distributed File Cache Consistency Gray and Cheriton By Farid Merchant Date: 9/21/05.
Advanced Eager Scheduling for Java-Based Adaptively Parallel Computing Michael O. Neary & Peter Cappello Computer Science Department UC Santa Barbara.
VICTORIA UNIVERSITY OF WELLINGTON Te Whare Wananga o te Upoko o te Ika a Maui SWEN 432 Advanced Database Design and Implementation Partitioning and Replication.
1 The Theory of NP-Completeness 2 Cook ’ s Theorem (1971) Prof. Cook Toronto U. Receiving Turing Award (1982) Discussing difficult problems: worst case.
Lecture 3: Uninformed Search
1 Branch and Bound Searching Strategies Updated: 12/27/2010.
FATCOP: A Mixed Integer Program Solver Michael FerrisQun Chen Department of Computer Sciences University of Wisconsin-Madison Jeff Linderoth, Argonne.
CS 584. Discrete Optimization Problems A discrete optimization problem can be expressed as (S, f) S is the set of all feasible solutions f is the cost.
Internet-Based TSP Computation with Javelin++ Michael Neary & Peter Cappello Computer Science, UCSB.
Branch and Bound Algorithms Present by Tina Yang Qianmei Feng.
1 Fault-Tolerant Mechanism for Hierarchical Branch and Bound Algorithm Université A/Mira de Béjaïa CEntre de Recherche sur l’Information Scientifique et.
FATCOP: A Mixed Integer Program Solver Michael FerrisQun Chen University of Wisconsin-Madison Jeffrey Linderoth Argonne National Laboratories.
© The McGraw-Hill Companies, Inc., Chapter 12 On-Line Algorithms.
1 Ch.19 Divide and Conquer. 2 BIRD’S-EYE VIEW Divide and conquer algorithms Decompose a problem instance into several smaller independent instances May.
CSCE350 Algorithms and Data Structure Lecture 21 Jianjun Hu Department of Computer Science and Engineering University of South Carolina
Data Mining By Farzana Forhad CS 157B. Agenda Decision Tree and ID3 Rough Set Theory Clustering.
Using Ant Agents to Combine Reactive and Proactive strategies for Routing in Mobile Ad Hoc Networks Fredrick Ducatelle, Gianni di caro, and Luca Maria.
11 -1 Chapter 12 On-Line Algorithms On-Line Algorithms On-line algorithms are used to solve on-line problems. The disk scheduling problem The requests.
Branch and Bound Searching Strategies
Java-Based Parallel Computing on the Internet: Javelin 2.0 & Beyond Michael Neary & Peter Cappello Computer Science, UCSB.
1 The Theory of NP-Completeness 2 Review: Finding lower bound by problem transformation Problem X reduces to problem Y (X  Y ) iff X can be solved by.
Lecture 3 – MapReduce: Implementation CSE 490h – Introduction to Distributed Computing, Spring 2009 Except as otherwise noted, the content of this presentation.
Hard Problems Some problems are hard to solve.
Backtracking And Branch And Bound
Real-Time Ray Tracing Stefan Popov.
Analysis and design of algorithm
Rui Wu, Jose Painumkal, Sergiu M. Dascalu, Frederick C. Harris, Jr
Analysis & Design of Algorithms (CSCE 321)
Branch and Bound Example
Adaptivity and Dynamic Load Balancing
Presentation transcript:

A Grid-enabled Branch and Bound Algorithm for Solving Challenging Combinatorial Optimization Problems Authors: M. Mezmaz, N. Melab and E-G. Talbi Presented by Saad Alhowaimel

Table of Content Introduction The Proposed Approach: Concepts and Operators The Proposed Grid-enabled Parallel B&B Experimentation

Introduction The problem is NP-hard and complex. The Branch and Bound reduce the computation time, but not exploring time. B&B components: – branching – Bounding – selection – elimination Parallelism on gird: – load balancing – fault tolerance – termination detection – global information sharing

The Proposed Approach: Concepts and Operators Construct a permutation tree and explore it using a DFS strategy List of active nodes. – Explored but not yet visited – Cover all the nodes which can be potentially explored from each node of the list. – Algorithm stops once the list of active nodes becomes empty – Used for exploration. Interval – Optimize communication – Check-pointing operations Fold operator – Deduces an interval from a list of active nodes Unfold operator. – Deduces a list of active nodes from an interval

The Proposed Approach: Concepts and Operators Node weight:

The Proposed Approach: Concepts and Operators Node Number:

The Proposed Approach: Concepts and Operators Node range:

The Proposed Approach: Concepts and Operators Fold Operator: – deduces an interval from a list of active nodes

The Proposed Approach: Concepts and Operators Fold Operator:

The Proposed Approach: Concepts and Operators Unfold Operator: – Deduces a list of active node form an interval.

The Proposed Grid-enabled Parallel B&B Farmer-Worker Approach: – Farmer One processor play the role of frame Keep a copy of all unexplored intervals – Worker All other processors. Each process explores an Interval Both worker and farmer mange the best solution found so far

The Proposed Grid-enabled Parallel B&B Fault Tolerance: – The coordinator manages a possible failure of the farmer by periodically saving in tow files. – INTERVALS: It has a copy of all the intervals. – SOLUTION: It has the global best solution. – The Updating process Let [A,B[ be an interval being explored in a B&B process, and [A’,B’[ its copy in INTERVALS Exploring increase A and branching decrease B’.

The Proposed Grid-enabled Parallel B&B Load Balance: – The coordinator assign an interval to B&B process using two operator: Selection operator: – select an interval from INTERVALS Partitioning operator: – divide an interval [A,B[ to [A,C[ and [C,B[. – The holder process keeps[A,C[ and requesting process keeps [C,B[ – The partitioning point C depend on the power and availability of the processors. – At the beginning C is equal to A. – Any interval that smaller than parameterized threshold is duplicated which speed up the search in termination phase.

The Proposed Grid-enabled Parallel B&B Termination Detection: – At the beginning INTERVALS has only one interval. – During exploration the number of B&B process is equal to the number of interval while its size continuously decreases. – The INTERVALS size is the sum of all the length of its interval. – Any empty interval (the beginning is higher than the end) is removed automatically. – The algorithm stop once the INTERVALS is empty.

The Proposed Grid-enabled Parallel B&B Solution Sharing: – The solution sharing is done by using SOLUTION. – It contains the global optimal solution. – Each B&B process manages its own local optimal solution. – The objective: is to reflect any improvement of any local optimal solution in all the other B&B processes. A B&B process initializes its local optimal solution by SOLUTION. Immediately informs the coordinator each time its local solution is improved Regularly reads SOLUTION to update its local optimal solution.

Experimentation The permutation flow-shop problem. – One of the important problem in industrial areas. – the problem is – Where: The position of the machines is fixed. A machine can handle only one job at a time.

Experimentation The experimentation on grid. – A problems of 50 jobs on 20 machines (Ta0561) – This instance has never been solved optimally. – The best known solution of Ta056 has a cost of 3681 with (near-optimal resolution method). – It is made up of 1889 processors belonging to 9 clusters. – Three clusters belong to three different education departments of University de Lille. – Six clusters belong Grid’5000

Experimentation Experiment Result. – Two runs have been performed, and the optimal solution is found with a cost equal to – First experiment: The upper bound set with 3681 average of processors 500 with a peak of 1245 processors. Experiment time One month and three weeks – Second experiment: The upper bound set with 3680 average of processors 328 with a peak of 1195 processors. Experiment time is 25 days. More than 2 million check pointing operations About 6 billion nodes were explored. a cumulative computation time of about 22 years..

Questions?