Hypergraph‐Based Dynamic Partitioning and Load Balancing

Çatalyürek, Ümit V.; Bozda¢g, Doruk; Boman, Erik G.; Devine, Karen Dragon; Heaphy, Robert; Riesen, Lee Ann

doi:10.1002/9780470558027.ch15

Cited by 4 publications

(5 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This model represents the current state of the application with a hypergraph where some vertices are fixed to partitions. Existing parallel hypergraph partitioning tools could not handle fixed--vertices, hence we have developed multi--level parallel hypergraph partitioning with fixed--vertices using Zoltan framework [1,2,3].…”

Section: Static and Dynamic Load Balancingmentioning

confidence: 99%

SciDAC Institute: Combinatorial Scientific Computing and Petascale Simulations (CSCAPES) (Final Report)

Çatalyürek¹

2014

View full text Add to dashboard Cite

Section: Static and Dynamic Load Balancingmentioning

confidence: 99%

SciDAC Institute: Combinatorial Scientific Computing and Petascale Simulations (CSCAPES) (Final Report)

Çatalyürek¹

2014

View full text Add to dashboard Cite

“…And a partitioning objective defined over the weighted hyperedges is optimized while a partitioning constraint is maintained. Our partitioning algorithm is based on the multilevel approach [8][9][10], which naturally fall into three main phases. Firstly, the coarsening phase constructs a sequence of successively smaller hypergraphs.…”

Section: Parallel Schemes Based On Hypergraph Partitioningmentioning

confidence: 99%

Parallel Computing of Laser Propulsion with Hypergraph Partitioning

Zeng

Zhao²,

Wang

2013

AMR

View full text Add to dashboard Cite

As one of the most important methods to study laser propelled rocket, the numerical simulation of laser propulsion has drawn an ever increasing attention at present. Nevertheless, the traditional serial computing cannot satisfy the practical requirements because of high calculation precision, insatiable memory overhead and considerable computation time. In this paper, we study on a general algorithm for laser propulsion, and divide the computing domain by using a multilevel hypergraph partitioning algorithm. Furthermore, MPI allreduce, overlapping communication with computation and non blocking communication are adopted to decrease the communication time when dealing with global communication. Finally, parallel performance about two typical configurations on a China-made supercomputer shows the smallest value of speedup ratio is more than 123 when the number of processors is 256. In conclusion, our parallel method is effective and practical in numerical simulation of laser propulsion.

show abstract

“…In this study, we use five partitioning algorithms: recursive coordinate bisection (RCB) [15], RCB with a post re-mapping step (RCB+Remap), AdaptiveRepart [22], HyperGraph [9], and Hilbert space-filling curve (HSFC). While RCB, RCB+Remap, HyperGraph, and HSFC are implemented in Zoltan [15], AdaptiveRepart is a part of the ParMetis library [18] to which Zoltan has an interface.…”

Section: Partitioning Algorithmsmentioning

confidence: 99%

“…Related work includes the AdaptiveRepart algorithm [22] in the ParMetis library [18], the VPI scheme [31] in the Jostle package [30], and the new hypergraph partitioner [9] in the Zoltan Library [15]. A common characteristic for these three bodies of work, is that they focus on developing state-of-the-art algorithms with the ability to efficiently optimize a number of metrics given some user-defined parameters.…”

Section: Introductionmentioning

confidence: 99%

The Benefits of Adaptive Partitioning for Parallel AMR Applications

Steensland

2008

View full text Add to dashboard Cite

Parallel adaptive mesh refinement methods potentially lead to realistic modeling of complex three-dimensional physical phenomena. However, the dynamics inherent in these methods present significant challenges in data partitioning and load balancing. Significant human resources, including time, effort, experience, and knowledge, are required for determining the optimal partitioning technique for each new simulation. In reality, scientists resort to using the on-board partitioner of the computational framework, or to using the partitioning industry standard, ParMetis. Adaptive partitioning refers to repeatedly selecting, configuring and invoking the optimal partitioning technique at run-time, based on the current state of the computer and application. In theory, adaptive partitioning automatically delivers superior performance and eliminates the need for repeatedly spending valuable human resources for determining the optimal static partitioning technique. In practice, however, enabling frameworks are non-existent due to the inherent significant inter-disciplinary research challenges. This paper presents a study of a simple implementation of adaptive partitioning and discusses implied potential benefits from the perspective of common groups of users within computational science. The study is based on a large set of data derived from experiments including six real-life, multi-time-step adaptive applications from various scientific domains, five complementing and fundamentally different partitioning techniques, a large set of parameters corresponding to a wide spectrum of computing environments, and a flexible cost function that considers the relative impact of multiple partitioning metrics and diverse partitioning objectives. The results show that even a simple implementation of adaptive partitioning can automatically generate results statistically equivalent to the best static partitioning. Thus, it is possible

show abstract

Hypergraph‐Based Dynamic Partitioning and Load Balancing

Cited by 4 publications

References 31 publications

SciDAC Institute: Combinatorial Scientific Computing and Petascale Simulations (CSCAPES) (Final Report)

SciDAC Institute: Combinatorial Scientific Computing and Petascale Simulations (CSCAPES) (Final Report)

Parallel Computing of Laser Propulsion with Hypergraph Partitioning

The Benefits of Adaptive Partitioning for Parallel AMR Applications

Contact Info

Product

Resources

About