Grid’BnB: A Parallel Branch and Bound Framework for Grids

Caromel, Denis; Costanzo, Alexandre; Baduel, Laurent; Matsuoka, Satoshi

doi:10.1007/978-3-540-77220-0_51

Cited by 14 publications

(6 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It has been recently a method of choice for experimenting the solving of optimization problems using Grid computing, because few data has to be exchanged between nodes [1]. Another implementation, described in [7], uses several hundreds of nodes of the Grid'5000 platform. Good speedups are achieved up to a few hundreds of processors but, interestingly, their conclusion is that the execution time tends to stabilize afterwards.…”

Section: Introductionmentioning

confidence: 99%

Experiments in Parallel Constraint-Based Local Search

Caniou¹,

Codognet²,

Díaz³

et al. 2011

Evolutionary Computation in Combinatorial Optimization

View full text Add to dashboard Cite

Abstract. We present a parallel implementation of a constraint-based local search algorithm and investigate its performance results on hardware with several hundreds of processors. We choose as basic constraint solving algorithm for these experiments the "adaptive search" method, an efficient sequential local search method for Constraint Satisfaction Problems. The implemented algorithm is a parallel version of adaptive search in a multiple independent-walk manner, that is, each process is an independent search engine and there is no communication between the simultaneous computations. Preliminary performance evaluation on a variety of classical CSPs benchmarks shows that speedups are very good for a few tens of processors, and good up to a few hundreds of processors.

show abstract

Section: Introductionmentioning

confidence: 99%

Experiments in Parallel Constraint-Based Local Search

Caniou¹,

Codognet²,

Díaz³

et al. 2011

Evolutionary Computation in Combinatorial Optimization

View full text Add to dashboard Cite

show abstract

“…In the domain of combinatorial optimization, the most popular complete method that has been parallelized at a large scale is the classical branch and bound method [26], because it does not require much information to be communicated between parallel processes: basically only the current upper (or lower) bound of the solution. It has thus been a method of choice for experimenting the solving of optimization problems using grid computing, see for instance [1] and also [17], which use several hundreds of nodes of the Grid'5000 platform. Good speed-ups are achieved up to a few hundreds of processors but interestingly, their conclusion is that runtimes tend to stabilize afterward.…”

Section: Parallel Complete Solversmentioning

confidence: 99%

Estimating parallel runtimes for randomized algorithms in constraint solving

et al. 2015

View full text Add to dashboard Cite

International audienceThis paper presents a detailed analysis of the scalability and par-allelization of Local Search algorithms for constraint-based and SAT (Boolean satisfiability) solvers. We propose a framework to estimate the parallel performance of a given algorithm by analyzing the runtime behavior of its sequential version. Indeed, by approximating the runtime distribution of the sequential process with statistical methods, the runtime behavior of the parallel process can be predicted by a model based on order statistics. We apply this approach to study the parallel performance of a Constraint-Based Local Search solver (Adaptive Search), two SAT Local Search solvers (namely Sparrow and CCASAT), and a propagation-based constraint solver (Gecode, with a random labeling heuristic). We compare the performance predicted by our model to actual parallel implementations of those methods using up to 384 processes. We show that the model is accurate and predicts performance close to the empirical data. Moreover, as we study different types of problems, we observe that the experimented solvers exhibit different behaviors and that their runtime distributions can be approximated by two types of distributions: exponential (shifted and non-shifted) and lognormal. Our results show that the proposed framework estimates the runtime of the parallel algorithm with an average discrepancy of 21% w.r.t. the empirical data across all the experiments with the maximum allowed number of processors for each technique

show abstract

“…It has been recently a method of choice for experimenting the solving of optimization problems using Grid computing, because few data has to be exchanged between nodes [13]. Another implementation, described in [14], uses several hundreds of nodes of the GRID'5000 platform. Good speedups are achieved up to a few hundreds of processors but, interestingly, their conclusion is that the execution time tends to stabilize afterwards even if global information is exchanged.…”

Section: Introductionmentioning

confidence: 99%

Communication in Parallel Algorithms for Constraint-Based Local Search

Caniou¹,

Codognet

2011

2011 IEEE International Symposium on Parallel and Distributed Processing Workshops and PHD Forum

View full text Add to dashboard Cite

We address the issue of parallelizing constraint solvers based on local search methods for massively parallel architectures, involving several thousands of CPUs. We present a family of a constraint-based local search algorithms and investigate their performance results on hardwares with several hundreds of processors. The first method is a basic independent multiple-walk algorithm: each processor runs a local search starting from a distinct initial configuration and the first one which will reach a solution will notify the others and stop all computations. These simple methods have good performances, and good speedups can be achieved up to a few hundreds of processors. Then we consider 2 versions with communication between processors: 1) every c iterations, each processor sends the current value (cost) of its configuration to others and a processor who received a better cost from another processor can decide to stop its current search with a probability p; 2) the number of iterations corresponding to the cost is also transfered. Both the received cost and the number of iterations have to be better for a processor to decide to draw a probability and restart.Several experiments involving more than 100 processors have been conducted and different values of p have been tried to consider more or less "autistic" processors. However results show that it is very difficult to achieve better performance than the initial method without communication.

show abstract

Grid’BnB: A Parallel Branch and Bound Framework for Grids

Cited by 14 publications

References 16 publications

Experiments in Parallel Constraint-Based Local Search

Experiments in Parallel Constraint-Based Local Search

Estimating parallel runtimes for randomized algorithms in constraint solving

Communication in Parallel Algorithms for Constraint-Based Local Search

Contact Info

Product

Resources

About