中国科学院大学计算机学院徐俊刚,10月30日学术报告信息(中国科学院大学 徐俊刚 )...

10月30日学术报告信息(中国科学院大学 徐俊刚 )

2015年10月27日16时

人评论

次

报告题目：Performance Analysis on Big Data Platform and Applications

报告日期及时间：2015.10.30 上午10:00-11:00

报告地点： 计算机学院B404

报告人：徐俊刚 教授 (Prof. Jungang Xu)

报告人国籍： 中国

报告人单位：中国科学院大学(University of Chinese Academy of Sciences)

报告人简介：

Jungang Xu is a professor of School of Computer and Control Engineering (SCCE) at University of Chinese Academy of

Sciences (UCAS). He is IEEE and ACM member, CCF senior member. He joined SCCE at UCAS as a faculty member in

2005, received his BS and MS degrees from Southwest Petroleum University (SWPU) and Shandong University (SDU) in

1996 and 1999 respectively, and received his PhD degree from Institute of Software, Chinese Academy of Sciences in 2003.

During 2003-2005, he pursued post-doctor research in Tsinghua University (THU). His current research interests include cloud, big data management and man-like intelligence. His research has been supported in part by the National Natural Science Foundation of China, the National

Key Technology R&D Program of China, etc.

报告摘要：

In recent years, the development of new economy and Internet enables the rapid growth of global data volume. IDC expects that

global data volume will reach 44ZB by 2020, which brings a tremendous challenge to the traditional computational frameworks.

Big data computational frameworks, such as Hadoop, Spark, are becoming important platforms to handle big data. However,

due to unreasonable configuration of big data platforms and improper design of big data applications, it is difficult to achieve

theoretical peak performance of the system. In this topic, we focus on performance analysis on big data platform and

applications. First, we introduce common performance profiling and analysis technologies for big data system; Second, we

present a multi-level performance model of big data system that covers the main software stack levels of Hadoop. Meanwhile, according to the performance bottleneck position, the performance bottlenecks are divided into four categories including nodes,

tasks, stages and user code; Third, we propose several performance analytical methods for big data system. Fourth, we

introduce a performance profile and analysis tool --BDPA (Big Data System Performance Analysis) that we implemented. Finally, we summarize the topic and introduce the future work.

邀请人： 龚奕利 副教授