Scaling up MapReduce-based Big Data Processing on Multi-GPU systems

Jiang, Hai; Chen, Yi; Qiao, Zhi; Weng, Tien‐Hsiung; Li, Kuan‐Ching

doi:10.1007/s10586-014-0400-1

Cited by 71 publications

(20 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Such computational paradigms have been preferred due to their reduced costs and inherent scalability, which pose many challenges to scalable systems and applications in terms of information access, storage and retrieval. Cluster computing [2], cloud computing technology [3], data and knowledge bases, distributed information retrieval technology [4] and networking technology [5] should all converge to address the scalability concern. Furthermore, with the advent of emerging computing architectures (e.g., SMTs, GPUs, and multicores), the importance of designing techniques explicitly targeting these systems is becoming more and more important.…”

Section: Editorialmentioning

confidence: 99%

Special Issue Editorial: Big Data for Mobile Services

Jung

2018

Mobile Netw Appl

View full text Add to dashboard Cite

We are in era of "Big Data". As data and knowledge volume keep increasing while global means for information dissemination continue to diversify, new methods, modeling paradigms and structures are needed to efficiently mount scalability requirements [1]. In the last few years, we have seen the proliferation of the use of heterogeneous distributed systems, ranging from simple Networks of Workstations, to highly complex grid computing environments. Such computational paradigms have been preferred due to their reduced costs and inherent scalability, which pose many challenges to scalable systems and applications in terms of information access, storage and retrieval. Cluster computing [2], cloud computing technology [3], data and knowledge bases, distributed information retrieval technology [4] and networking technology [5] should all converge to address the scalability concern. Furthermore, with the advent of emerging computing architectures (e.g., SMTs, GPUs, and multicores), the importance of designing techniques explicitly targeting these systems is becoming more and more important.The 8th EAI International Conference on Big Data Technologies and Applications (BDTA) has mainly focused on a wide array of scalability issues and investigated new approaches to tackle problems arising from the ever-growing size and complexity of information of all kinds.This special issue features six selected papers with high quality from BDTA held in Gwangju, Korea, November 23-24, 2017. The first paper, entitled "Fine-grained Big Traffic Data Reverse-charge System: a Method of Saving Expenses", presents an interesting framework to process big traffic data. Particularly, by analyzing the data streams from mobile applications, they have experimentally employed the proposed reverse charge system, and participants' responses to questionnaires indicate a win-win situation for customers as well as online enterprises.The second paper "Owner-Borrower Model for Recommenders in O2O Services" considers the O2O recommendation framework for matchmaking between owner and borrower. Especially, they propose a recommendation method based on applying a tensor factorization method and the Gale-Shapley algorithm. The tensor factorization is used for estimating preferences of the owners and the borrowers. The proposed framework can be expected to implement various O2O services.The third paper "Detecting Human Emotions in a Large Size of Database by using Ensemble Classification Model" presents a novel human emotion detection system by using ensemble classification model (with Support Vector Machine, Neural Network and k-Nearest Neighbors). The evaluation results show that emotion classification by using the ensemble classification method can be better than the single classificationbased model.The fourth paper "Transportation Object Detection with Bag of Visual Words Model by PLSA and MLP" investigates how to analyze visual big data for transportation services. It shows a new visual detection method for transportation based on probabilistic latent seman...

show abstract

Section: Editorialmentioning

confidence: 99%

Special Issue Editorial: Big Data for Mobile Services

Jung

2018

Mobile Netw Appl

View full text Add to dashboard Cite

show abstract

“…Finally, at the mere system-side, several works focus the attention on very interesting applicative settings, such as the case of supporting MapReduce-based big data processing on multi-GPU systems (e.g., [55]), and the case of efficiently supporting GIS polygon overlay computation with MapReduce for spatial big data processing (e.g., [56]). …”

Section: Mapreduce Algorithms For Big Data Processingmentioning

confidence: 99%

An Effective and Efficient MapReduce Algorithm for Computing BFS-Based Traversals of Large-Scale RDF Graphs

2016

View full text Add to dashboard Cite

Nowadays, a leading instance of big data is represented by Web data that lead to the definition of so-called big Web data. Indeed, extending beyond to a large number of critical applications (e.g., Web advertisement), these data expose several characteristics that clearly adhere to the well-known 3V properties (i.e., volume, velocity, variety). Resource Description Framework (RDF) is a significant formalism and language for the so-called Semantic Web, due to the fact that a very wide family of Web entities can be naturally modeled in a graph-shaped manner. In this context, RDF graphs play a first-class role, because they are widely used in the context of modern Web applications and systems, including the emerging context of social networks. When RDF graphs are defined on top of big (Web) data, they lead to the so-called large-scale RDF graphs, which reasonably populate the next-generation Semantic Web. In order to process such kind of big data, MapReduce, an open source computational framework specifically tailored to big data processing, has emerged during the last years as the reference implementation for this critical setting. In line with this trend, in this paper, we present an approach for efficiently implementing traversals of large-scale RDF graphs over MapReduce that is based on the Breadth First Search (BFS) strategy for visiting (RDF) graphs to be decomposed and processed according to the MapReduce framework. We demonstrate how such implementation speeds-up the analysis of RDF graphs with respect to competitor approaches. Experimental results clearly support our contributions.

show abstract

Section: Editorialmentioning

confidence: 99%

Special Issue Editorial: Advances on Scalable Information Systems for Big Data (InfoScale 2014)

Jung

2015

Mobile Netw Appl

View full text Add to dashboard Cite

EditorialWe are in era of BBig Data^. As data and knowledge volume keep increasing while global means for information dissemination continue to diversify, new methods, modeling paradigms and structures are needed to efficiently mount scalability requirements [1]. In the last few years, we have seen the proliferation of the use of heterogeneous distributed systems, ranging from simple Networks of Workstations, to highly complex grid computing environments. Such computational paradigms have been preferred due to their reduced costs and inherent scalability, which pose many challenges to scalable systems and applications in terms of information access, storage and retrieval. Cluster computing [2], cloud computing technology [3], data and knowledge bases, distributed information retrieval technology [4] and networking technology [5] should all converge to address the scalability concern. Furthermore, with the advent of emerging computing architectures (e.g., SMTs, GPUs, and multicores), the importance of designing techniques explicitly targeting these systems is becoming more and more important.The 5th International Conference on Scalable Information Systems (InfoScale) has mainly focused on a wide array of scalability issues and investigated new approaches to tackle problems arising from the ever-growing size and complexity of information of all kinds.This special issue features six selected papers with high quality from InfoScale held in Seoul, Korea, September 25-26, 2014. The first paper, entitled BReal-time Event Detection on Social Data Stream^, presents an interesting framework to process big social data. Particularly, by analyzing the data streams from social media, useful events can be detected and applied to provide users with intelligent context-aware services.The second paper BSimilarity Searching for the Big Dataâ uthored by Pavel Zezula considers the scalable framework for big data processing. The paper firstly finds out a necessity to discover descriptive information of complex and heterogeneous objects to make them accessible. Second, multimodal search structures are requested to efficiently conduct complex similarity queries possibly in outsourced environments while preserving privacy.The third paper BWeighted Similarity Schemes for High Scalability in User-Based Collaborative Filtering^by Pirasteh et al. presents a novel recommender system (RecSys) based on similarity integration method. Different from conventional RecSys (which is based on single similarity between a pair of users), the proposed system collects all possible similarities between two users, and integrates them for representing the relationship between them in the best manners.The fourth paper BA Novel Ranking Model for a Large-Scale Scientific Publication^investigates how to rank academic publications (e.g., research papers, conference proceedings, and journals), which are modeled in very highly complex networks. Given a large amount of publication data, authors have proposed multi-layered network model (call N-star model) to proces...

show abstract

Scaling up MapReduce-based Big Data Processing on Multi-GPU systems

Cited by 71 publications

References 7 publications

Special Issue Editorial: Big Data for Mobile Services

Special Issue Editorial: Big Data for Mobile Services

An Effective and Efficient MapReduce Algorithm for Computing BFS-Based Traversals of Large-Scale RDF Graphs

Special Issue Editorial: Advances on Scalable Information Systems for Big Data (InfoScale 2014)

Contact Info

Product

Resources

About