Benchmarking in the Data Center

Workshop Scope and Topics

While traditional cloud and high performance computing workloads are quite different, many of the components used in high performance computers are also used in traditional data centers. Furthermore, big data and high performance computing are converging, both in terms of hardware and software used. It is therefore helpful to have a common subset of performance metrics that are useful in assessing data center efficiency. Of particular interest, is that more than half of the computers on the Top500 list are in Asia.

Survey

Please fill in the survey here Randomly selected participants will receive:

Date and Location


Wednesday 16 January

Vanburgh Hotel, Guangzhou, China

Workshop Program

Time Name Affilation Title Slides
13:00-13:30 Guo Liang Open Data Center Committee & China Academy of Information and Communications Technology An overview of data centers in China
13:30-14:00 Robert Henschel Indiana University & SPEC Using the SPEC HPG Benchmarks for Better Analysis and Evaluation of Current and Future HPC Systems slides
14:00-14:10 Vikcy Xie Open Data Center Committee & China Academy of Information and Communications Technology Liquid cooling technology and performance efficiency in data center
14:10-14:30 Yao Xiao Shenzhen University The Role of Privacy in Benchmarking Geo-Distributed Data Centers
14:30-15:00 Wu-chun Feng Virginia Tech The Green500: Metrics, Methodologies, and Workloads
15:00-15:30 Break
15:30-16:00 Piotr Bala University of Warsaw Benchmarking Java as a possibility to merge HPC and Big Data processing slides
16:00-16:30 Xiaoyi Lu Ohio State University Benchmarking and Accelerating Big Data and Deep Learning Systems on Modern HPC Architecture slides
16:30-17:00 Zhuowei Si Intel Intel MPI benchmarks slides
17:00-17:25 Juan Chen National University of Defense Technology Moderated Discussion
17:25-17:30 Closing

Submissions

The workshop solicits: Papers will be published and presented. Previous work will be briefly introduced. Both will form the basis of discussion groups during the conference workshop.

The first portion of the workshop will then be devoted to the selected position papers, introduction of key points from abstracts submitted for discussion and a brief survey overview. There will then be breakout discussions, followed by a wrap-up discussion. A post workshop edited collection of papers on sustainable and effective benchmarking of converged high performance and big data computing infrastructures is also planned. This will be comprised of extended workshop papers, invited papers and summary papers with new viewpoints resulting from the workshop.


Both the conference proceedings and the post workshop volume will have open access options. For information about ACM open access options, please see here.

ACM templates can be found here

Please make submissions here.

Schedule

Registration

Please register at the main conference website, here

.

Contact

datacenter.benchmark.workshop@parallel.computer