Volume 15 Number 2
April 2018
Article Contents
Chao-Long Zhang, Yuan-Ping Xu, Zhi-Jie Xu, Jia He, Jing Wang and Jian-Hua Adu. A Fuzzy Neural Network Based Dynamic Data Allocation Model on Heterogeneous Multi-GPUs for Large-scale Computations. International Journal of Automation and Computing, vol. 15, no. 2, pp. 181-193, 2018. https://doi.org/10.1007/s11633-018-1120-4
Citation: Chao-Long Zhang, Yuan-Ping Xu, Zhi-Jie Xu, Jia He, Jing Wang and Jian-Hua Adu. A Fuzzy Neural Network Based Dynamic Data Allocation Model on Heterogeneous Multi-GPUs for Large-scale Computations. International Journal of Automation and Computing, vol. 15, no. 2, pp. 181-193, 2018. https://doi.org/10.1007/s11633-018-1120-4

A Fuzzy Neural Network Based Dynamic Data Allocation Model on Heterogeneous Multi-GPUs for Large-scale Computations

Author Biography:
• Chao-Long Zhang received the B. Eng. and M. Sc. degrees in software engineering from Chengdu University of Information Technology, China in 2014 and 2017, respectively. He is currently a Ph. D. degree candidate with School of Computing and Engineering, University of Huddersfield, UK.
His research interests include highperformance computing (HPC), computer vision, and deep learning network applications.
E-mail:chaolong.zhang@hud.ac.uk
ORCID iD:0000-0003-4990-4636

Zhi-Jie Xu received the B. Eng. degree in communication engineering from the Xi'an University of Science and Technology, China in 1991. After graduation, he first started as an electronics engineer before moving to the United Kingdom and worked as a research scientist in the Robotics Lab at the University of Derby. He received the Ph. D. degree in 2000 from the University of Derby based on his research work in virtual reality-based manufacturing simulation and robotics systems. He has been employed as a full time academic member of staff since April 1999 serving the roles of lecturer, senior lecturer, reader and professor respectively at the University of Huddersfield in UK. He has published over one hundred peer-reviewed journal and conference papers as well as edited 5 books in the relevant fields. He has successfully supervised 8 Ph. D. students to completion while securing substantial research and industrial grants. He is a member of the IEEE, IET, BCS, and a fellow of HEA, and editors for multiple prestigious academic journals and conferences. He is the current President of the Chinese Automation and Computing Society in the United Kingdom.
His research interests include visualization, HCI, vision systems, and machine learning.
E-mail:z.xu@hud.ac.uk

Jia He received B. Eng. and M. sc. degrees in computer science and technology from Southwest Normal University of China, China in 1989 and 1996, respectively, and received Ph. D. degree in computer science from University of Electronic Science and Technology of China, China in 2012. She is currently a professor and the Dean with School of Computer Science, Chengdu University of Information Technology, China.
Her research interests include computer vision, artificial intelligence, and pattern recognition.
E-mail:hejia@cuit.edu.cn

Jing Wang received the Ph. D. degree from University of Huddersfield, UK in 2012. He worked as a research fellow and carried out independent research work on image processing, analysing and understanding in University of Huddersfield, UK before 2017. He is now working at Sheffield Hallam University as a lecturer in software engineering and computer science.
His research interest is real-world applications of computer vision systems.
E-mail:jing.wang@shu.ac.uk

Jian-Hua Adu received B. Sc. degree in applied physics from Minzu University of China, China in 1999, received M. Sc. degree in computer science from Shandong University, China in 2006, and received Ph. D. degree in computer science from Sichuan University, China in 2012. He is currently an associate professor with School of Software Engineering, Chengdu University of Information Technology, China.
His research interests include image fusion and segmentation, medical image processing and analysis, and pattern recognition.

• Corresponding author: Yuan-Ping Xu received the B. Eng. degree in computer science and technology from Southwest Jiaotong University, China in 2003, and M. Sc. and Ph. D. degrees in software engineering from University of Huddersfield, UK in 2004 and 2009, respectively. From February 2009 to November 2010, he worked as a research fellow in the Centre of Precision Technologies, University of Huddersfield, UK. He is currently a professor with School of Software Engineering, Chengdu University of Information Technology, China.
His research interests include knowledge-based systems, expert systems, big data analysis and deep learning network applications.
E-mail:ypxu@cuit.edu.cn (Corresponding author)
ORCID iD:0000-0002-4536-6220
• Accepted: 2018-02-09
• Published Online: 2018-03-12
Fund Project:

National Natural Science Foundation of China 61203172

the SSTP of Sichuan 2018YYJC0994

the SSTP of Sichuan 2017JY0011

Shenzhen STPP GJHZ20160301164521358

• The parallel computation capabilities of modern graphics processing units (GPUs) have attracted increasing attention from researchers and engineers who have been conducting high computational throughput studies. However, current single GPU based engineering solutions are often struggling to fulfill their real-time requirements. Thus, the multi-GPU-based approach has become a popular and cost-effective choice for tackling the demands. In those cases, the computational load balancing over multiple GPU "nodes" is often the key and bottleneck that affect the quality and performance of the real-time system. The existing load balancing approaches are mainly based on the assumption that all GPU nodes in the same computer framework are of equal computational performance, which is often not the case due to cluster design and other legacy issues. This paper presents a novel dynamic load balancing (DLB) model for rapid data division and allocation on heterogeneous GPU nodes based on an innovative fuzzy neural network (FNN). In this research, a 5-state parameter feedback mechanism defining the overall cluster and node performance is proposed. The corresponding FNN-based DLB model will be capable of monitoring and predicting individual node performance under different workload scenarios. A real-time adaptive scheduler has been devised to reorganize the data inputs to each node when necessary to maintain their runtime computational performance. The devised model has been implemented on two dimensional (2D) discrete wavelet transform (DWT) applications for evaluation. Experiment results show that this DLB model enables a high computational throughput while ensuring real-time and precision requirements from complex computational tasks.
通讯作者: 陈斌, bchen63@163.com
• 1.

沈阳化工大学材料科学与工程学院 沈阳 110142

Figures (9)  / Tables (4)

A Fuzzy Neural Network Based Dynamic Data Allocation Model on Heterogeneous Multi-GPUs for Large-scale Computations

• Corresponding author:Yuan-Ping Xu received the B. Eng. degree in computer science and technology from Southwest Jiaotong University, China in 2003, and M. Sc. and Ph. D. degrees in software engineering from University of Huddersfield, UK in 2004 and 2009, respectively. From February 2009 to November 2010, he worked as a research fellow in the Centre of Precision Technologies, University of Huddersfield, UK. He is currently a professor with School of Software Engineering, Chengdu University of Information Technology, China.     His research interests include knowledge-based systems, expert systems, big data analysis and deep learning network applications.     E-mail:ypxu@cuit.edu.cn (Corresponding author)     ORCID iD:0000-0002-4536-6220
Fund Project:

National Natural Science Foundation of China 61203172

the SSTP of Sichuan 2018YYJC0994

the SSTP of Sichuan 2017JY0011

Shenzhen STPP GJHZ20160301164521358

Abstract: The parallel computation capabilities of modern graphics processing units (GPUs) have attracted increasing attention from researchers and engineers who have been conducting high computational throughput studies. However, current single GPU based engineering solutions are often struggling to fulfill their real-time requirements. Thus, the multi-GPU-based approach has become a popular and cost-effective choice for tackling the demands. In those cases, the computational load balancing over multiple GPU "nodes" is often the key and bottleneck that affect the quality and performance of the real-time system. The existing load balancing approaches are mainly based on the assumption that all GPU nodes in the same computer framework are of equal computational performance, which is often not the case due to cluster design and other legacy issues. This paper presents a novel dynamic load balancing (DLB) model for rapid data division and allocation on heterogeneous GPU nodes based on an innovative fuzzy neural network (FNN). In this research, a 5-state parameter feedback mechanism defining the overall cluster and node performance is proposed. The corresponding FNN-based DLB model will be capable of monitoring and predicting individual node performance under different workload scenarios. A real-time adaptive scheduler has been devised to reorganize the data inputs to each node when necessary to maintain their runtime computational performance. The devised model has been implemented on two dimensional (2D) discrete wavelet transform (DWT) applications for evaluation. Experiment results show that this DLB model enables a high computational throughput while ensuring real-time and precision requirements from complex computational tasks.

Chao-Long Zhang, Yuan-Ping Xu, Zhi-Jie Xu, Jia He, Jing Wang and Jian-Hua Adu. A Fuzzy Neural Network Based Dynamic Data Allocation Model on Heterogeneous Multi-GPUs for Large-scale Computations. International Journal of Automation and Computing, vol. 15, no. 2, pp. 181-193, 2018. https://doi.org/10.1007/s11633-018-1120-4
 Citation: Chao-Long Zhang, Yuan-Ping Xu, Zhi-Jie Xu, Jia He, Jing Wang and Jian-Hua Adu. A Fuzzy Neural Network Based Dynamic Data Allocation Model on Heterogeneous Multi-GPUs for Large-scale Computations. International Journal of Automation and Computing, vol. 15, no. 2, pp. 181-193, 2018. https://doi.org/10.1007/s11633-018-1120-4
Reference (38)

/