Dynamic Data Management Among Multiple Databases for Optimization of Parallel Computations in Heterogeneous HPC Systems

Rościszewski, Paweł

doi:10.5121/csit.2014.4709

Cited by 2 publications

(4 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…What is more, it is possible to combine the functionality of multiple optimizers. For example, in , we present a data prefetching optimizer that uses an internal optimizer for scheduling.…”

Section: Parallelization and Optimization Of Computations In Kernelhivementioning

confidence: 99%

“…The programmer only designs a workflow application in a graphical editor, uses OpenCL to code computational kernels that are assigned to workflow nodes, and provides input data. Depending on the optimization goal chosen, for example, minimization of the execution time considered in this work, a version using data prefetching or minimization of the execution time with a bound on the power consumption , the system selects compute devices on which computations are to be run, determines optimal OpenCL grid configurations, and determines optimal data partitioning to obtain good granularity for good load balancing and consequently good speed‐ups on the given number of compute devices. Then the actual run is performed in a distributed and possibly heterogeneous setting with both CPUs and GPUs.…”

Section: Summary and Future Workmentioning

confidence: 99%

See 1 more Smart Citation

KernelHive: a new workflow‐based framework for multilevel high performance computing using clusters and workstations with CPUs and GPUs

Roźciszewski

Czarnul

Lewandowski

et al. 2015

Concurrency and Computation

View full text Add to dashboard Cite

SUMMARYThe paper presents a new open-source framework called KernelHive for multilevel parallelization of computations among various clusters, cluster nodes, and finally, among both CPUs and GPUs for a particular application. An application is modeled as an acyclic directed graph with a possibility to run nodes in parallel and automatic expansion of nodes (called node unrolling) depending on the number of computation units available. A methodology is proposed for parallelization and mapping of an application to the environment that includes selection of devices using a chosen optimizer, selection of best grid configurations for compute devices, optimization of data partitioning and the execution. One of possibly many scheduling algorithms can be selected considering execution time, power consumption, and so on. An easy-to-use GUI is provided for modeling and monitoring with a repository of ready-to-use constructs and computational kernels. The methodology, execution times, and scalability have been demonstrated for a distributed and parallel password-breaking example run in a heterogeneous environment with a cluster and servers with different numbers of nodes and both CPUs and GPUs. Additionally, performance of the framework has been compared with an MPI + OpenCL implementation using a parallel geospatial interpolation application employing up to 40 cluster nodes and 320 cores.

show abstract

“…What is more, it is possible to combine the functionality of multiple optimizers. For example, in , we present a data prefetching optimizer that uses an internal optimizer for scheduling.…”

Section: Parallelization and Optimization Of Computations In Kernelhivementioning

confidence: 99%

Section: Summary and Future Workmentioning

confidence: 99%

KernelHive: a new workflow‐based framework for multilevel high performance computing using clusters and workstations with CPUs and GPUs

Roźciszewski

Czarnul

Lewandowski

et al. 2015

Concurrency and Computation

View full text Add to dashboard Cite

show abstract

“…The results of the experiments with prefetching optimizer presented in [14] revealed, that in case of data packages of similar size and efficient data management, network is constantly loaded. The benefits of prefetching in a more complex environment ( Figure 8) are significantly smaller than in case of a single device ( Figure 6).…”

Section: Need For a Network-aware Scheduling Schemementioning

confidence: 99%

“…We presented the KernelHive system and its performance capabilities in [4] and proposed an execution optimizer focusing on energy efficiency in [5]. This paper is an extended version of [14], where we added data intensity capabilities to the KernelHive system. For this purpose we proposed MongoDB [6] database as a backend.…”

Section: Introductionmentioning

confidence: 99%

Network-Aware Data Prefetching Optimization of Computations in a Heterogeneous HPC Framework

Rościszewski¹

2014

IJCNC

Self Cite

View full text Add to dashboard Cite

Rapid development of diverse computer architectures and hardware accelerators caused that designing parallel systems faces new problems resulting from their heterogeneity. Our implementation of a parallel system called KernelHive allows to efficiently run applications in a heterogeneous environment consisting of multiple collections of nodes with different types of computing devices. The execution engine of the system is open for optimizer implementations, focusing on various criteria. In this paper, we propose a new optimizer for KernelHive, that utilizes distributed databases and performs data prefetching to optimize the execution time of applications, which process large input data. Employing a versatile data management scheme, which allows combining various distributed data providers, we propose using NoSQL databases for our purposes. We support our solution with results of experiments with real executions of our OpenCL implementation of a regular expression matching application in various hardware configurations. Additionally, we propose a network-aware scheduling scheme for selecting hardware for the proposed optimizer and present simulations that demonstrate its advantages.

show abstract

Dynamic Data Management Among Multiple Databases for Optimization of Parallel Computations in Heterogeneous HPC Systems

Abstract: ABSTRACT

Cited by 2 publications

References 9 publications

KernelHive: a new workflow‐based framework for multilevel high performance computing using clusters and workstations with CPUs and GPUs

KernelHive: a new workflow‐based framework for multilevel high performance computing using clusters and workstations with CPUs and GPUs

Network-Aware Data Prefetching Optimization of Computations in a Heterogeneous HPC Framework

Contact Info

Product

Resources

About