An empirical study of data decomposition for software parallelization

Meade, Anne; Deeptimahanti, Deva Kumar; Buckley, Jim; Collins, John J.

doi:10.1016/j.jss.2016.02.002

Cited by 9 publications

(3 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, these resources may define the granularity level that the system can support [20]. There have been few empirical studies for performing data decomposition in the HPC field, [21] investigated this approach when designing parallel applications. Additionally, the state-ofpractice was studied with probing tools used to perform this function.…”

Section: Related Workmentioning

confidence: 99%

A Proposed Architecture for Parallel HPC-based Resource Management System for Big Data Applications

Shehri¹,

Khemakhem²,

Basuhail³

et al. 2019

Adv. sci. technol. eng. syst. j.

View full text Add to dashboard Cite

Big data can be considered to be at the forefront of the present and future research activities. The volume of data needing to be processed is growing dramatically in both velocity and variety. In response, many big data technologies have emerged to tackle the challenges of collecting, processing and storing such large-scale datasets. Highperformance computing (HPC) is a technology that is used to perform computations as fast as possible. This is achieved by integrating heterogeneous hardware and crafting software and algorithms to exploit the parallelism provided by HPC. The performance capabilities afforded by HPC have made it an attractive environment for supporting scientific workflows and big data computing. This has led to a convergence of the HPC and big data fields. However, big data applications usually do not fully exploit the performance available in HPC clusters. This is so due to such applications being written in high-level programming languages and do not provide support for exploiting parallelism as do other parallel programming models. The objective of this research paper is to enhance the performance of big data applications on HPC clusters without sacrificing the power consumption of HPC. This can be achieved by building a parallel HPC-based Resource Management System to exploit the capabilities of HPC resources efficiently.

show abstract

Section: Related Workmentioning

confidence: 99%

A Proposed Architecture for Parallel HPC-based Resource Management System for Big Data Applications

Shehri¹,

Khemakhem²,

Basuhail³

et al. 2019

Adv. sci. technol. eng. syst. j.

View full text Add to dashboard Cite

show abstract

“…All such factors introduce new challenges to the process of designing and optimizing parallel applications for these systems. Indeed, in parallelization, the design of data partitioning and distribution is one of the most time-and effort-consuming tasks in runtime optimization [4], [5]. Data partitioning is an important composition unit of data access patterns and significantly affects the performance of an application [6].…”

Section: Introductionmentioning

confidence: 99%

A Proposed Data Partitioning Approach on Heterogeneous HPC Platforms: Data Locality Perspective

Al-Hashimi

Basuhail

2021

IEEE Access

View full text Add to dashboard Cite

We propose a new data partitioning approach to improve the performance of heterogeneous parallel applications in modern high-performance computing (HPC) systems. Existing approaches do not consider an important aspect that has a critical impact on the performance of parallel applications: the method of assigning partitions to each processor so as to minimize the communication cost and hence minimize data movement, which dominates energy and performance cost. Such an aspect for managing data locality is important for a large range of applications. Therefore, to achieve efficient data partitioning, we propose a method for distribution considering this aspect. Our algorithm seeks to minimize execution time by using two models. The first is a fine-grained computational model of heterogeneous processors, which is sufficiently adequate and accurate to guarantee efficient partitioning results that maximize utilization. The second is a communication model of heterogeneous processors to minimize data motion and hide communication overheads. The correctness of our algorithm was analyzed and validated. The complexity of our algorithm is approximately of order ( × + × ), where is (where steps is the step size between data points in the computational model of each processor), and is the number of heterogeneous processors. The experiments were performed on AZIZ supercomputer using two types of applications: an application with no dependency between its partitions, i.e., matrix multiplication, and another one with high dependency between its partitions, i.e., the Jacobi method. The results show the efficiency of our algorithm in improving performance.

show abstract

“…Execution model that is being used: Compiled code for CPU Q3: TYPICAL USER PROFILES FOR THE LANGUAGE Roles of the users of this language: End-user Technical knowledge required: Languages (Goal language) Q4: EFFECTIVENESS OF THE LANGUAGE -> Success evaluated, Quantitative comparison performed, Explicit comparison of the language proposal with respect to distinct settings/context/configurations, Two test cases. They performed a number of tests regarding the scalability of the system with respect to a number of factors, Metrics (Lines of code, Satisfaction, Time), Impact on the productivity gains brought (Learnability, Lower cognitive overload, Easier to remember, Expressiveness, Easier to use -Qualitative), Products' performance gains brought (Computation efficiency, Scalability -Quantitative)A.2.6 Java[18,19,20,21,22] Q2: NATURE OF THE LANGUAGEApplication Domain: Grid w applications to Ray tracing and Sequencing; Machine Learning; Specify policies to transform divide and conquer sequential programs into parallel executions Purpose of the language: Formalization of the requirements of the problem, Formalization of the solution, Simulation of the solution, Implement the solution, Data Interpretation Key Advantages: Performance, Portability, Easiness of configuration, Orchestration and Usability (Effectiveness/Efficiency/Satisfaction) Paradigms underlying the language: Object-Oriented, Hybrid (Language to schedule constraint solving) There is a concrete syntax for the language and the preferred representation type is Textual Tool support for the language: Interpreters, Compilers Technologies used to create the language tool suite: XML based technology ( A XML like syntax to describe classes and methods to be scheduled ) Execution stack requirements to support the artifacts created with those languages: VM Supervisor (JVM on grid), OS (any), IO architecture (Grid), Libraries (Apache Spark, 77 Weka 3.6.0, Hadoop 0.20) Execution model that is being used: Virtual Execution Environment ( Java Virtual Machine ), Distributed middleware (Hadoop, Apache Spark), HPC Libraries (Apache Spark), Bytecode for virtual machine (JVM on Grid) Q3: TYPICAL USER PROFILES FOR THE LANGUAGE Roles of the users of this language: End-user Technical knowledge required: Languages (Java) Q4: EFFECTIVENESS OF THE LANGUAGE -> S uccess evaluated, Quantitative comparison performed, Metrics (Lines of code, Time), Impact on the productivity gains brought (Easier to use, Compact representation), Products' performance gains brought (Computation efficiency, Scalability -quantitative) A.2.7 OpenCL [27, 28] Q2: NATURE OF THE LANGUAGE Application Domain: CFD (any application that benefits from GPU), Big Data processing Purpose of the language: Formalization of the requirements of the problem, Implement the solution Key advantages: Performance, Portability, Easiness of configuration, Orchestration, Usability (Effectiveness/Efficiency/Satisfaction) Paradigms underlying the language: Object-Oriented There is a concrete syntax for the language and the preferred representation type can be both Textual and Diagrammatic Existing tool support for the language: Compilers, Tool suite Technologies used to create the language tool suite: GenERTiCA source code generator Execution stack requirements to support the artifacts created with those languages: multiple OS supported Execution model being used: Distributed middleware, HPC Libraries, Bytecode for virtual machine, Compiled code for CPU, Compiled code for GPU The language target specific hardware and GPUs or multi-core architecturesQ3: TYPICAL USER PROFILES FOR THE LANGUAGE Roles of the users of this language: End-user Technical knowledge required: ...…”

mentioning

confidence: 99%

Programming languages for data-Intensive HPC applications: A systematic mapping study

et al. 2020

View full text Add to dashboard Cite

We present the results of a systematic literature review that examines the main paradigms and properties of programming languages developed for and used in High Performance Computing for Big Data processing. The systematic literature review is based on a combination of automated keyword-based search in the Elsevier Science Direct database and further digital databases for articles published in international peer-reviewed journals and conferences, leading to an initial sample of 420 articles, which was then narrowed down in a second phase to 152 articles found relevant and published 2006-2018. The manual analysis of these articles allowed us to identify 26 languages used in 33 of these articles for HPC for Big Data processing. We analyzed the languages and their usage in these articles by 22 criteria and summarize the results in this article. We evaluate the outcomes of the literature review by comparing them with opinions of domain experts. Our results indicate that, for instance, the majority of the used HPC languages in the context of Big Data are text-based general-purpose programming languages and target the end-user community.

show abstract

An empirical study of data decomposition for software parallelization

Cited by 9 publications

References 27 publications

A Proposed Architecture for Parallel HPC-based Resource Management System for Big Data Applications

A Proposed Architecture for Parallel HPC-based Resource Management System for Big Data Applications

A Proposed Data Partitioning Approach on Heterogeneous HPC Platforms: Data Locality Perspective

Programming languages for data-Intensive HPC applications: A systematic mapping study

Contact Info

Product

Resources

About