We achieved the first positions in the world rankings of the Graph500 and the Green Graph500 Benchmarks.
Both the tenth Graph500 and the fifth Green Graph500 lists were announced at the ISC2015 in Frankfurt, Germany. We won the first positions in the world rankings of the Graph500 and the Green Graph500 Benchmarks. These results show that our algorithms and implementations are efficient and scalable for big data computation in each computing environment; distributed-memory supercomputers, shared-memory supercomputers, Intel/AMD servers, and Android devices.
◯ Graph500 List
We have achieved 38621.4 GTEPS for Kronecker graph with SCALE 40 on 82,944 (of 88,128) nodes of the K computer by our new MPI-based implementation. We investigated the bottlenecks in our implementation and improved a performance of communications and memory accesses. This entry is almost twice faster than Sequoia, which achieves 23,751 GTEPS for SCALE 41 on 82,944 nodes and 663,552 cores and the first position in previous list. And we also updated the TEPS scores on distributed-memory supercomputers TSUBAME 2.5, as 1344.95 GTEPS for SCALE 36. On the other side, on shared-memory computers, our OpenMP-based implementation achieves 174.704 GTEPS on two racks of UV 2000. This score indicates the fastest of single-node systems and the highest TEPS per core in an upper position than this. In addition, we have obtained a highly efficient result on the HP Superdome X.
|1||K computer||RIKEN Advanced Institute for Computational Science (AICS)||82,944||663,552||40||38621.4|
|12||TSUBAME 2.5||GSIC, Tokyo Institute of Technology||1,024||12,288||36||1344.95|
|42||ismuv2k2 (SGI UV 2000)||The Institute of Statistical Mathematics||1||1,280||33||174.704|
|45||HP Integrity Superdome X||HP Ft. Collins||1||240||33||128.336|
The certification of Graph500 (July 13, 2015)
◯ Green Graph500 list
＊ Big Data Category (more than or equal to SCALE 30)We defended the first position of the big data category in the fifth Green Graph500 list. All the first, second, third, fourth, and fifth entries in this list are our results. The #1, #2, and #3 ware obtained by our OpenMP-based implementation on each a fat memory node. And we updated 62.93 MTEPS/W and 31.3 GTEPS and defended the first position. Furthermore, the entries #4, #5 and #6 are obtained by using an offloading technique for NVM (Non-volatile memory) on a single node. This implementation can handle large problems, which exceed on-board memory.
|4||44.35||Tokyo Institute of Technology||EBD-RH2288V3||32||14.86|
|5||35.87||Tokyo Institute of Technology||GraphCREST-Custom #1||32||10.64|
|6||28.88||Tokyo Institute of Technology||MEM-CREST Node #2||30||7.98|
The certification in Big data category (July 13, 2015)
＊ Small Data Category (less than SCALE 30)No significant changed in this list. The highest TEPS/W entry is same as last. Our entries located on the 2nd to 9th ranks.
|1||445.92||George Washington University||Colonial||20||122.18|
|4||204.38||Tokyo Institute of Technology||EBD-GoldenBox-Prototype||21||1.64|
◯ Presentation slides
Prof. Fujisawa talked the 10th Graph500 and the 5th Green Graph500 lists at the HPC in Asia Session in ISC15. The presentation slides are available at the following URL. Title: A Challenge to Graph500 Benchmark: Trillion-Scale Graph Processing on K Computer Author: Katsuki Fujisawa Co-authors: Koji Ueno, Hitoshi Sato, Yuichiro Yasui, Keita Iwabuchi, Toyotaro Suzumura, and Ryo Mizote
◯ Graph500 and Green Graph500 Collaborative Research Team
- ・Katsuki Fujisawa (Research Director) / Institute of Mathematics for Industry, Kyushu University
- ・Toyotaro Suzumura / University College Dublin
- ・Hitoshi Sato / Global Scientific Information and Computing Center, Tokyo Institute of Technology
- ・Koji Ueno / Global Scientific Information and Computing Center, Tokyo Institute of Technology
- ・Yuichiro Yasui / Center for Co-evolutional Social Systems, Kyushu University
- ・Ryu Mizote / Global Scientific Information and Computing Center, Tokyo Institute of Technology
- ・Keita Iwabuchi / Global Scientific Information and Computing Center, Tokyo Institute of Technology