CENG 505 Parallel Computing I Parallel Computing for

  • Slides: 14
Download presentation
CENG 505 Parallel Computing I Parallel Computing for Scheduling Algorithms PREPARED BY: Atıl KURT

CENG 505 Parallel Computing I Parallel Computing for Scheduling Algorithms PREPARED BY: Atıl KURT TO: Cem ÖZDOĞAN 05. 01. 11 1 ATIL KURT

CENG 505 PARALLEL COMPUTING OUTLINE Ø Introduction Ø Scheduling Problems Ø Optimum Algorithms Ø

CENG 505 PARALLEL COMPUTING OUTLINE Ø Introduction Ø Scheduling Problems Ø Optimum Algorithms Ø Heuristic algorithms Ø Scheduling for Parallel Processing Ø Conclusion 05. 01. 11 2 ATIL KURT

CENG 505 PARALLEL COMPUTING INTRODUCTION v Scheduling : Establishing the timing of the use

CENG 505 PARALLEL COMPUTING INTRODUCTION v Scheduling : Establishing the timing of the use of equipment, facilities, and human activities v. Assignment v. Sequencing v Scheduling problems shown by α/β/γ symbols v α: Machine environment v β: Task system v γ: Optimality criterion 05. 01. 11 3 ATIL KURT

CENG 505 PARALLEL COMPUTING Scheduling Problems α β γ 1 – Single machine Pmtn

CENG 505 PARALLEL COMPUTING Scheduling Problems α β γ 1 – Single machine Pmtn – Preemptive Cmax – maximum completion time P – Parallel machine Chain – Precedence constraint Lmax – maximum lateness Q – Uniform parallel machine NR – Non resumable ∑Cj – Total completion time R – Unrelated parallel machine rj – ready times ∑Tj – Total tardiness F – Flow shop Tool-change ∑uj – total tardy jobs O – Open shop Eligibility ∑wj. Cj – Weighted completion J – job sjop Deterioting times R/tool-change, /Cmax : Makespan minimization problem on unrelated parallel machines with tool change constraints 05. 01. 11 4 ATIL KURT

CENG 505 PARALLEL COMPUTING Optimum Algorithms v Optimum Special Algorithms Ø Moore algorithm 1//∑uj

CENG 505 PARALLEL COMPUTING Optimum Algorithms v Optimum Special Algorithms Ø Moore algorithm 1//∑uj Ø SPT 1//∑Cj , P//∑Cj Ø EDD 1//Lmax v Mathematical model Solution tools Ø Branch and Bound Algorithms o Kouiki, S. , Jemni, M. Ladhari, T. , 2010, Design of parallel distributed algorithm for the permutation flow shop Processor 1 2 4 6 Tail 092 2560 1400 750 560 Tail 098 15050 6940 3860 2870 Run times (ms) 5 05. 01. 11 ATIL KURT

CENG 505 PARALLEL COMPUTING Heuristic Algorithms v Special Heuristics v LPT list algorithms P//Cmax

CENG 505 PARALLEL COMPUTING Heuristic Algorithms v Special Heuristics v LPT list algorithms P//Cmax v CDS heuristic F//Cmax v Metaheuristics Local Search methods A top-level general strategy which guides other heuristics to search for feasible solutions in domains where the task is hard. v Genetic Algorithm v Simulated Annealing v Tabu search v Ant Colony optimiziation 05. 01. 11 6 ATIL KURT

CENG 505 PARALLEL COMPUTING Heuristic Algorithms (cont’d) v Genetic algorithm Ø Mimics the process

CENG 505 PARALLEL COMPUTING Heuristic Algorithms (cont’d) v Genetic algorithm Ø Mimics the process of natural selection Ø The Algorithms is generally as follows: 1. Randomly generate an initial population M(0) 2. Compute and save the fitness u(m) for each individual m in the current population M(t) 3. Define selection probabilities p(m) for each individual m in M(t) so that p(m) is proportional to u(m) 4. Generate M(t+1) by probabilistically selecting individuals from M(t) to produce offspring via genetic operators 5. Repeat step 2 until satisfying solution is obtained. 05. 01. 11 7 ATIL KURT

CENG 505 PARALLEL COMPUTING Heuristic Algorithms (cont’d) Genetic algorithm (articles) v. Wang, N. ,

CENG 505 PARALLEL COMPUTING Heuristic Algorithms (cont’d) Genetic algorithm (articles) v. Wang, N. , (2004) A parallel computing application of the genetic algorithm for lubrication optimization. Tribology Letters, Vol. 18 v. Alba, E. , Troye J. M. , A survey of parallel distributed genetic algorithms v. Bozejko, W. Wodecki, M. , Parallel genetic algorithm for the flow shop scheduling problems 05. 01. 11 8 ATIL KURT

CENG 505 PARALLEL COMPUTING Heuristic Algorithms (cont’d) v Simulated Annealing Ø Analogy of solid

CENG 505 PARALLEL COMPUTING Heuristic Algorithms (cont’d) v Simulated Annealing Ø Analogy of solid thermodynamic state evaluation simulation Ø Czech , J. Z. , Czarnas, P. Parallel simulated annealing for the vehicle routing problem with time windows Ø Barry, D. A. , Morris, J. , Parallel simulated annealing using CILK language: Aplication in estimating transport parameters for groundwater contaminants 05. 01. 11 9 ATIL KURT

CENG 505 PARALLEL COMPUTING Heuristic Algorithms (cont’d) v Tabu Search Ø Uses a strategy

CENG 505 PARALLEL COMPUTING Heuristic Algorithms (cont’d) v Tabu Search Ø Uses a strategy of proccessing in the direction of biggest decrease or the smallest increase, of the objective function. Ø Gendreau, M. , laporte, G. , Semet, F. (2000). A dynamic model and parallel tabu search heuristic for real-time ambulance relocation. ØYi, H. , Yuhui, Q. , Guangyuan, L. , Kaiyou, L. , (2003). A parallel tabu search approach based on genetic crossover operation. ØCraınıc, T. G. , Gendreau M. , Potvin , J. , Parallel tabu search 05. 01. 11 10 ATIL KURT

CENG 505 PARALLEL COMPUTING Heuristic Algorithms (cont’d) Ant Colony optimization v Seeks to mimic

CENG 505 PARALLEL COMPUTING Heuristic Algorithms (cont’d) Ant Colony optimization v Seeks to mimic an ant's apparent ability to find the shortest distance between two points. Number of Processors v Randall, M. , A parallel implementation of Ant Colony Optimization. v Delisle, P. , Krajecki, M. , Gravel, M. , Gagné, C. , Parallel implementation of an Ant Colony Optimization metaheurıstıc with OPENMP. 05. 01. 11 11 ATIL KURT

CENG 505 PARALLEL COMPUTING Scheduling for Parallel Processing Ø Scheduling algorithm can be used

CENG 505 PARALLEL COMPUTING Scheduling for Parallel Processing Ø Scheduling algorithm can be used for the Parallel Processing Ø Reduction in CPU with load balancing Ø Use Processors Machine Task Job CPU of Task Processing times Ø Processor are identical : Processor load balancing P//Cmax Ø Processor are not identical : Processor load balancing R//Cmax 05. 01. 11 12 ATIL KURT

CENG 505 PARALLEL COMPUTING Conclusion Ø Parallel Algorithms are applicable for Scheduling Problems Ø

CENG 505 PARALLEL COMPUTING Conclusion Ø Parallel Algorithms are applicable for Scheduling Problems Ø These algorithms have a lots of application in literature Ø These algorithms give good solution for the NP-hard scheduling algorithms Ø These Algorithms can be used for my thesis. Ø B & B algorithm Ø Metaheuristics 05. 01. 11 13 ATIL KURT

CENG 505 PARALLEL COMPUTING THANKS !!! FOR YOUR LISTENING 05. 01. 11 14 ATIL

CENG 505 PARALLEL COMPUTING THANKS !!! FOR YOUR LISTENING 05. 01. 11 14 ATIL KURT