Distributed Scheduling of Parallel Hybrid Computations

Agarwal, Sumeet; Narang, Ankur; Shyamasundar, R. K.

doi:10.1007/978-3-642-10631-6_115

Cited by 3 publications

(7 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We can see that due to skewed probability of choosing which balls go to which bin, the probability of successful attempts goes down compared to the case of uniform probability [10]. Even though we chose ring distance based probability variation, actual processor distance based probability variation can be similarly analyzed with suitable probability distribution.…”

Section: Proofmentioning

confidence: 93%

“…Due to space constraints at each place in the system, the algorithm needs to keep track of space availability at each worker and place to ensure physical deadlock freedom. Our distributed scheduling algorithm ensures distributed physical deadlock avoidance by using depth based ordering of computations for execution [10]. Physical deadlock avoidance ensures that the system always has space to guarantee the execution of a certain number of paths, that can vary during the execution of the computation DAG.…”

Section: Distributed Scheduling Of Hybrid Computation In Bounded Spacementioning

confidence: 99%

“…3D torus interconnect). This paper leverages the distributed deadlock avoidance strategy for deadlock free execution in [10] for efficient scheduling of hybrid parallel computations. [11] presents the results for bounded space scheduling of hybrid computations.…”

Section: Introductionmentioning

confidence: 99%

“…This paper leverages the distributed deadlock avoidance strategy for deadlock free execution in [10] for efficient scheduling of hybrid parallel computations. [11] presents the results for bounded space scheduling of hybrid computations. In this paper, we present detailed proofs and analysis of the time and message complexity of the unconstrained space scheduling and the bounded space scheduling algorithms for hybrid computations.…”

Section: Introductionmentioning

confidence: 99%

“…Further, it considered only a single processor per place whereas in reality each place typically has multiple processors per place such as an SMP configuration with multiple cores (2-core P6 chip from IBM [8], 8-core Niagara T2 chip from Sun [9]). [10] considers affinity driven distributed scheduling (P m |M j prec|C max ) of multi-place parallel computations with physical deadlock freedom (using distributed deadlock avoidance strategy). It provides time and message complexity lower and upper bounds and deadlock freedom proof but the algorithms and proofs are limited to affinity driven activities and it assumes same distances between places on the cluster.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Performance driven distributed scheduling of parallel hybrid computations

Narang

Shyamasundar

2011

Theoretical Computer Science

Self Cite

View full text Add to dashboard Cite

a b s t r a c tExascale computing is fast becoming a mainstream research area. In order to realize exascale performance, it is necessary to have efficient scheduling of large parallel computations with scalable performance on a large number of cores/processors. The scheduler needs to execute in a pure distributed and online fashion, should follow affinity inherent in the computation and must have low time and message complexity. Further, it should also avoid physical deadlocks due to bounded resources including space/memory per core. Simultaneous consideration of these factors makes affinity driven distributed scheduling particularly challenging. We attempt to address this challenge for hybrid parallel computations which contain tasks that have pre-specified affinity to a place and also tasks that can be mapped to any place in the system. Specifically, we address two scheduling problems of the type P m |M j , prec|C max . This paper presents online distributed scheduling algorithms for hybrid parallel computations assuming both unconstrained and bounded space per place. We also present the time and message complexity for distributed scheduling of hybrid computations. To the best of our knowledge, this is the first time that distributed scheduling algorithms for hybrid parallel computations have been presented and analyzed for time and message bounds under both unconstrained space and bounded space.

show abstract

Section: Proofmentioning

confidence: 93%

Section: Distributed Scheduling Of Hybrid Computation In Bounded Spacementioning

confidence: 99%