Optimal Scheduling Algorithms for Communication Constrained Parallel Processing

Altılar, D. Turgay; Paker, Yakup

doi:10.1007/3-540-45706-2_25

Cited by 30 publications

(24 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The first scenario is for the root processor to only distribute and collect load without computation. This is because we try to compare its performance with one in the previous papers [3,4]. Here we assume that our multi-port star network is homogeneous, which means all of the children processors are identical in terms of the processing speed.…”

Section: A System Modelingmentioning

confidence: 99%

“…For these two algorithms the authors [3,4] found both the maximum processing throughput and I/O utilization, and the optimal number of processors for each of algorithms under a bus architecture, using divisible load analysis [1,2]. However the algorithms have inherent limitations of sequential I/O communication, due to the use of a bus based architecture, on communication in terms of throughput and the optimal number of processors.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Efficient Parallel Video Processing Through Concurrent Communication on a Multi-port Star Network

Lim

Robertazzi

2006

2006 40th Annual Conference on Information Sciences and Systems

View full text Add to dashboard Cite

Abstract-Communication delay in a processor network is very critical to the throughput for parallel video processing. We propose a simultaneous distribution and collection method (SD) from the root processor to children processors via a multi-port switch network. For the proposed mechanism, we analyze the video encoding time and derive a closed-form solution for a star interconnection network topology. The results show that the total encoding time is significantly faster than the previous method, Parallel Interlaced (PI). In addition, we achieve scalability in terms of the number of processors, which means that as the number of processors increases over the optimal number of processors of PI, one continues to achieve much better performance.Index Terms-Parallel video scheduling, divisible load theory, concurrent communication, star network.

show abstract

Section: A System Modelingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Efficient Parallel Video Processing Through Concurrent Communication on a Multi-port Star Network

Lim

Robertazzi

2006

2006 40th Annual Conference on Information Sciences and Systems

View full text Add to dashboard Cite

show abstract

“…It ranges from data partitioning for parallel video processing [2] to finding the optimal number of processors in linear algebra algorithms [4]. More generally, many results have been produced by the divisible load theory for various network topologies (see [10] for an overview).…”

Section: Related Workmentioning

confidence: 99%

Load-balancing scatter operations for grid computing

2004

View full text Add to dashboard Cite

We present solutions to statically load-balance scatter operations in parallel codes run on grids. Our load-balancing strategy is based on the modification of the data distributions used in scatter operations. We study the replacement of scatter operations with parameterized scatters, allowing custom distributions of data. The paper presents: (1) a general algorithm which finds an optimal distribution of data across processors; (2) a quicker guaranteed heuristic relying on hypotheses on communications and computations; (3) a policy on the ordering of the processors. Experimental results with an MPI scientific code illustrate the benefits obtained from our load-balancing.

show abstract

“…This divisible load model has been widely studied in the last several years, and Divisible Load Theory (DLT) has been popularized by the landmark book written in 1996 by Bharadwaj, Ghose, Mani and Robertazzi [4]. DLT has been applied to a large spectrum of scientific problems, including linear algebra [6], image processing [11,14], video and multimedia broadcasting [1,2], database searching [5], biological pattern-matching [13], and the processing of large distributed files [17].…”

Section: Introductionmentioning

confidence: 99%

Scheduling multiple divisible loads on a linear processor network

Gallet

Robert

Vivien

2007

2007 International Conference on Parallel and Distributed Systems

View full text Add to dashboard Cite

Min, Veeravalli, and Barlas have recently proposed strategies to minimize the overall execution time of one or several divisible loads on a heterogeneous linear network, using one or more installments [18,19]. We show on a very simple example that their approach does not always produce a solution and that, when it does, the solution is often suboptimal. We also show how to find an optimal scheduling for any instance, once the number of installments per load is given. Then, we formally prove that any optimal schedule has an infinite number of installments under a linear cost model as the one assumed in [18,19]. Such a cost model cannot be used to design practical multi-installment strategies. Finally, through extensive simulations we confirmed that the best solution is always produced by the linear programming approach.

show abstract

Optimal Scheduling Algorithms for Communication Constrained Parallel Processing

Cited by 30 publications

References 7 publications

Efficient Parallel Video Processing Through Concurrent Communication on a Multi-port Star Network

Efficient Parallel Video Processing Through Concurrent Communication on a Multi-port Star Network

Load-balancing scatter operations for grid computing

Scheduling multiple divisible loads on a linear processor network

Contact Info

Product

Resources

About