The 15th Graph500 list – which ranks supercomputers based on how quickly they can build knowledge from massive-scale data sets – was released Nov. 15 at Supercomputing 2017 (SC17), with Japan’s K-Computer defending its position in the number-one spot several years in a row.

The Graph500 is recognized as a leading indicator of high-performance computing (HPC) development and investment globally and often reveals trends regarding new technologies used in the machines. It provides a benchmark standard to test a supercomputer's abilities to construct, search, and conduct edge-detection for undirected graphs. 

Georgia Tech School of Computational Science and Engineering Chair David Bader, Peter Kogge of the University of Notre Dame, Andrew Lumsdaine of Pacific Northwest National Laboratory, and Rich Murphy of Micron Technology Inc., head the Graph500 executive committee. This committee – along with an International Multidisciplinary Steering Committee that comprises 30 international HPC experts from academia, national laboratories, and industry – rank Graph500 machines based on the benchmark standards.

“The supercomputers measured are used to analyze big data for cybersecurity, medical informatics, social networks, data enrichment, and symbolic networks. The list is released twice a year to coincide with the two largest high-performance computing HPC conferences – Supercomputing and International Supercomputing Conference (ISC) – and it is the dominant international benchmark for analytics platforms,” said Bader.

The Graph500 executive committee presented the top three:

• RIKEN Advanced Institute for Computational Science (AICS)’s K Computer
• National Supercomputing Center in Wuxi’s Sunway TaihuLight
• Lawrence Livermore National Laboratory’s DOE/NNSA/LLNL Sequoia

In addition to the announcement of the top three and highlighting the top 10 entries, the session also provided a discussion of Kogge’s Summary of Graph500 Analytics, Kernel 2 Reference Implementation (SSSP) and Results, as well as the stream benchmark discussion and proposals.

“The entries are truly international with machines from the U.S., Japan, Russia, China, Canada representing the countries with the most entries. Many other countries are represented as well amongst this year’s participants,” said Bader during the announcement ceremony.

“Data intensive supercomputer applications are increasingly important workloads, especially for big data problems, but are ill-suited for most of today’s computing platforms. The Graph500 has demonstrated the challenges of even simple analytics and its participants have been extraordinary,” said Bader.

Since the first Graph500 list with nine machines at SC10 in Nov. 2010, Graph500 has grown drastically to 235 entries in this year’s 15th Graph500 list at SC17.

The 16th Graph500 list will be released at ISC, next June in Frankfurt, Germany.