The K computer Operations: Experiences and Statistics

Yamamoto, Keiji; Uno, Atsuya; Murai, Hitoshi; Tsukamoto, Toshiyuki; Shoji, Fumiyoshi; Matsui, Shuji; Sekizawa, Ryuichi; Sueyasu, Fumichika; Uchiyama, Hiroshi; Okamoto, Mitsuo; Ohgushi, Nobuo; Takashina, Katsutoshi; Wakabayashi, Daisuke; Taguchi, Yoshitaka; Yokokawa, Mitsuo

doi:10.1016/j.procs.2014.05.052

Cited by 26 publications

(14 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On the DKRZ cluster, a majority of the executed jobs uses less than 32 nodes. On the other hand, the data presented in [26] and [30] show that on other systems, the majority of jobs uses more than 128 nodes (HECToR and MareNostrum), more than 512 nodes (Jugene and Jubl), or even more than 1024 nodes (K computer). This implies that the cost of CPU overheatings in these systems would probably be much larger on average.…”

Section: Discussionmentioning

confidence: 99%

CPU Overheating Characterization in HPC Systems: A Case Study

Platini

Ropars

Pelletier

et al. 2018

2018 IEEE/ACM 8th Workshop on Fault Tolerance for HPC at eXtreme Scale (FTXS)

View full text Add to dashboard Cite

With the increase in size of supercomputers, also increases the number of abnormal events. Some of these events might lead to an application failure. Others might simply impact the system efficiency. CPU overheating is one such event that decreases the system efficiency: when a CPU overheats, it reduces its frequency. This paper studies the problem of CPU overheating in supercomputers. In a first part, we analyze data collected over one year on a supercomputer of the top500 list to understand under which conditions CPU overheating occurs. Our analysis show that overheating events are due to some specific applications. In a second part, we evaluate the impact of such overheating events on the performance of MPI applications. Using 6 representative HPC benchmarks, we show that for a majority of the applications, a frequency drop on one CPU impacts the execution time of distributed runs proportionally to the duration and to the extent of the frequency drop.

show abstract

Section: Discussionmentioning

confidence: 99%

CPU Overheating Characterization in HPC Systems: A Case Study

Platini

Ropars

Pelletier

et al. 2018

2018 IEEE/ACM 8th Workshop on Fault Tolerance for HPC at eXtreme Scale (FTXS)

View full text Add to dashboard Cite

show abstract

“…The system includes 82,944 compute nodes connected by Tofu high-speed interconnects. For more details, see (Yamamoto et al 2014). 6 https://github.com/hazem2410/SEPNET…”

Section: Endnotesmentioning

confidence: 99%

Tie-formation process within the communities of the Japanese production network: application of an exponential random graph model

et al. 2019

View full text Add to dashboard Cite

This paper studies the driving forces behind the formation of ties within the major communities in the Japanese nationwide network of production, which contains one million firms and five million links between suppliers ("upstream" firms) and customers ("downstream" firms). We apply the Infomap algorithm to reveal the hierarchical structure of the production network. At the second level of the hierarchy, we find a reasonable community resolution, where the community size distribution follows a power law decay. Then, we estimate the tie formation within 100 communities of different sizes. The studied model considers a large set of attributes, including both endogenous attributes (network motifs, e.g., stars and triangles) and exogenous attributes (economic variables, e.g., net sales and firm size). The estimation results show that the considered model converges and presents a high goodness of fit (GoF) for all communities. Moreover, it is found that the forces explaining the formation of links between suppliers and customers differ among communities. Some attributes, such as reciprocity, popularity, activity, location homophily, bank homophily and sales statistics, are common drivers of internal link formation for most of the studied communities. However, transitivity is rejected as a significant influencing factor for most communities, reflecting an absence of a sense of trust and reliability between firms with a common partner. Finally, we show that sector homophily does not serve as an obvious mechanism of partnership at the community level in the production network.

show abstract

“…The asynchronous, distributed file I/O system of the K computer minimizes waiting time for frequent output of the atmospheric variables during the simulation. The K computer is also highly resilient and has a lower machine failure rate (Yamamoto et al 2014) compared to other peta-scale supercomputers, which is very important for long-term climate simulations with a large number of computational nodes.…”

Section: Computational Aspects Backgroundmentioning

confidence: 99%

Outcomes and challenges of global high-resolution non-hydrostatic atmospheric simulations using the K computer

Satoh

Tomita

Yashiro

et al. 2017

Prog. in Earth and Planet. Sci.

View full text Add to dashboard Cite

This article reviews the major outcomes of a 5-year (2011-2016) project using the K computer to perform global numerical atmospheric simulations based on the non-hydrostatic icosahedral atmospheric model (NICAM). The K computer was made available to the public in September 2012 and was used as a primary resource for Japan's Strategic Programs for Innovative Research (SPIRE), an initiative to investigate five strategic research areas; the NICAM project fell under the research area of climate and weather simulation sciences. Combining NICAM with high-performance computing has created new opportunities in three areas of research: (1) higher resolution global simulations that produce more realistic representations of convective systems, (2) multi-member ensemble simulations that are able to perform extended-range forecasts 10-30 days in advance, and (3) multi-decadal simulations for climatology and variability. Before the K computer era, NICAM was used to demonstrate realistic simulations of intra-seasonal oscillations including the Madden-Julian oscillation (MJO), merely as a case study approach. Thanks to the big leap in computational performance of the K computer, we could greatly increase the number of cases of MJO events for numerical simulations, in addition to integrating time and horizontal resolution. We conclude that the high-resolution global non-hydrostatic model, as used in this five-year project, improves the ability to forecast intra-seasonal oscillations and associated tropical cyclogenesis compared with that of the relatively coarser operational models currently in use. The impacts of the sub-kilometer resolution simulation and the multi-decadal simulations using NICAM are also reviewed.

show abstract

The K computer Operations: Experiences and Statistics

Cited by 26 publications

References 6 publications

CPU Overheating Characterization in HPC Systems: A Case Study

CPU Overheating Characterization in HPC Systems: A Case Study

Tie-formation process within the communities of the Japanese production network: application of an exponential random graph model

Outcomes and challenges of global high-resolution non-hydrostatic atmospheric simulations using the K computer

Contact Info

Product

Resources

About