当サイトでは、より良いサービスを提供するため、クッキーを利用しています。クッキーの使用に同意いただける場合は「同意」ボタンをクリックし、クッキーポリシーについては「詳細を見る」をクリックしてください。詳しくは当サイトの サイトポリシー をご確認ください。

詳細を見る...
ログイン サインアップ
言語:

WEKO3

  • トップ
  • ランキング
To

Field does not validate



インデックスリンク

インデックスツリー

  • RootNode

メールアドレスを入力してください。

WEKO

One fine body…

WEKO

One fine body…

アイテム

{"_buckets": {"deposit": "cd52946d-2ace-4c67-88ea-102c4c7633c7"}, "_deposit": {"created_by": 7, "id": "6374", "owners": [7], "pid": {"revision_id": 0, "type": "depid", "value": "6374"}, "status": "published"}, "_oai": {"id": "oai:nied-repo.bosai.go.jp:00006374", "sets": []}, "author_link": [], "item_10001_biblio_info_7": {"attribute_name": "書誌情報", "attribute_value_mlt": [{"bibliographicIssueDates": {"bibliographicIssueDate": "2012-05-23", "bibliographicIssueDateType": "Issued"}, "bibliographicIssueNumber": "2-3", "bibliographicPageEnd": "155", "bibliographicPageStart": "147", "bibliographicVolumeNumber": "28", "bibliographic_titles": [{"bibliographic_title": "Computer Science - Research and Development", "bibliographic_titleLang": "ja"}, {"bibliographic_title": "COMPUTER SCIENCE-RESEARCH AND DEVELOPMENT", "bibliographic_titleLang": "en"}]}]}, "item_10001_description_5": {"attribute_name": "抄録", "attribute_value_mlt": [{"subitem_description": "This paper proposes the design of ultra scalable MPI collective communication for the K computer, which consists of 82,944 computing nodes and is the world\u0027s first system over 10 PFLOPS. The nodes are connected by a Tofu interconnect that introduces six dimensional mesh/torus topology. Existing MPI libraries, however, perform poorly on such a direct network system since they assume typical cluster environments. Thus, we design collective algorithms optimized for the K computer.On the design of the algorithms, we place importance on collision-freeness for long messages and low latency for short messages. The long-message algorithms use multiple RDMA network interfaces and consist of neighbor communication in order to gain high bandwidth and avoid message collisions. On the other hand, the short-message algorithms are designed to reduce software overhead, which comes from the number of relaying nodes. The evaluation results on up to 55,296 nodes of the K computer show the new implementation outperforms the existing one for long messages by a factor of 4 to 11 times. It also shows the short-message algorithms complement the long-message ones.", "subitem_description_language": "en", "subitem_description_type": "Other"}]}, "item_10001_publisher_8": {"attribute_name": "出版者", "attribute_value_mlt": [{"subitem_publisher": "Springer Science and Business Media LLC", "subitem_publisher_language": "ja"}, {"subitem_publisher": "SPRINGER HEIDELBERG", "subitem_publisher_language": "en"}]}, "item_10001_relation_14": {"attribute_name": "DOI", "attribute_value_mlt": [{"subitem_relation_type_id": {"subitem_relation_type_id_text": "10.1007/s00450-012-0211-7"}}]}, "item_10001_source_id_9": {"attribute_name": "ISSN", "attribute_value_mlt": [{"subitem_source_identifier": "1865-2042", "subitem_source_identifier_type": "EISSN"}]}, "item_creator": {"attribute_name": "著者", "attribute_type": "creator", "attribute_value_mlt": [{"creatorNames": [{"creatorName": "Tomoya Adachi", "creatorNameLang": "ja"}, {"creatorName": "Tomoya Adachi", "creatorNameLang": "en"}]}, {"creatorNames": [{"creatorName": "Naoyuki Shida", "creatorNameLang": "ja"}, {"creatorName": "Naoyuki Shida", "creatorNameLang": "en"}]}, {"creatorNames": [{"creatorName": "Kenichi Miura", "creatorNameLang": "ja"}, {"creatorName": "Kenichi Miura", "creatorNameLang": "en"}]}, {"creatorNames": [{"creatorName": "Shinji Sumimoto", "creatorNameLang": "ja"}, {"creatorName": "Shinji Sumimoto", "creatorNameLang": "en"}]}, {"creatorNames": [{"creatorName": "Atsuya Uno", "creatorNameLang": "ja"}, {"creatorName": "Atsuya Uno", "creatorNameLang": "en"}]}, {"creatorNames": [{"creatorName": "Motoyoshi Kurokawa", "creatorNameLang": "ja"}, {"creatorName": "Motoyoshi Kurokawa", "creatorNameLang": "en"}]}, {"creatorNames": [{"creatorName": "Fumiyoshi Shoji", "creatorNameLang": "ja"}, {"creatorName": "Fumiyoshi Shoji", "creatorNameLang": "en"}]}, {"creatorNames": [{"creatorName": "Mitsuo Yokokawa", "creatorNameLang": "ja"}, {"creatorName": "Mitsuo Yokokawa", "creatorNameLang": "en"}]}]}, "item_language": {"attribute_name": "言語", "attribute_value_mlt": [{"subitem_language": "eng"}]}, "item_title": "The design of ultra scalable MPI collective communication on the K computer", "item_titles": {"attribute_name": "タイトル", "attribute_value_mlt": [{"subitem_title": "The design of ultra scalable MPI collective communication on the K computer", "subitem_title_language": "ja"}, {"subitem_title": "The design of ultra scalable MPI collective communication on the K computer", "subitem_title_language": "en"}]}, "item_type_id": "40001", "owner": "7", "path": ["1670839190650"], "permalink_uri": "https://nied-repo.bosai.go.jp/records/6374", "pubdate": {"attribute_name": "PubDate", "attribute_value": "2023-09-20"}, "publish_date": "2023-09-20", "publish_status": "0", "recid": "6374", "relation": {}, "relation_version_is_last": true, "title": ["The design of ultra scalable MPI collective communication on the K computer"], "weko_shared_id": -1}
  1. 防災科研関係論文

The design of ultra scalable MPI collective communication on the K computer

https://nied-repo.bosai.go.jp/records/6374
https://nied-repo.bosai.go.jp/records/6374
6a8ae682-39e4-41c2-9f20-79a91bca201d
Item type researchmap(1)
公開日 2023-09-20
タイトル
言語 ja
タイトル The design of ultra scalable MPI collective communication on the K computer
タイトル
言語 en
タイトル The design of ultra scalable MPI collective communication on the K computer
言語
言語 eng
著者 Tomoya Adachi

× Tomoya Adachi

ja Tomoya Adachi

en Tomoya Adachi

Search repository
Naoyuki Shida

× Naoyuki Shida

ja Naoyuki Shida

en Naoyuki Shida

Search repository
Kenichi Miura

× Kenichi Miura

ja Kenichi Miura

en Kenichi Miura

Search repository
Shinji Sumimoto

× Shinji Sumimoto

ja Shinji Sumimoto

en Shinji Sumimoto

Search repository
Atsuya Uno

× Atsuya Uno

ja Atsuya Uno

en Atsuya Uno

Search repository
Motoyoshi Kurokawa

× Motoyoshi Kurokawa

ja Motoyoshi Kurokawa

en Motoyoshi Kurokawa

Search repository
Fumiyoshi Shoji

× Fumiyoshi Shoji

ja Fumiyoshi Shoji

en Fumiyoshi Shoji

Search repository
Mitsuo Yokokawa

× Mitsuo Yokokawa

ja Mitsuo Yokokawa

en Mitsuo Yokokawa

Search repository
抄録
内容記述タイプ Other
内容記述 This paper proposes the design of ultra scalable MPI collective communication for the K computer, which consists of 82,944 computing nodes and is the world's first system over 10 PFLOPS. The nodes are connected by a Tofu interconnect that introduces six dimensional mesh/torus topology. Existing MPI libraries, however, perform poorly on such a direct network system since they assume typical cluster environments. Thus, we design collective algorithms optimized for the K computer.On the design of the algorithms, we place importance on collision-freeness for long messages and low latency for short messages. The long-message algorithms use multiple RDMA network interfaces and consist of neighbor communication in order to gain high bandwidth and avoid message collisions. On the other hand, the short-message algorithms are designed to reduce software overhead, which comes from the number of relaying nodes. The evaluation results on up to 55,296 nodes of the K computer show the new implementation outperforms the existing one for long messages by a factor of 4 to 11 times. It also shows the short-message algorithms complement the long-message ones.
言語 en
書誌情報 ja : Computer Science - Research and Development
en : COMPUTER SCIENCE-RESEARCH AND DEVELOPMENT

巻 28, 号 2-3, p. 147-155, 発行日 2012-05-23
出版者
言語 ja
出版者 Springer Science and Business Media LLC
出版者
言語 en
出版者 SPRINGER HEIDELBERG
ISSN
収録物識別子タイプ EISSN
収録物識別子 1865-2042
DOI
関連識別子 10.1007/s00450-012-0211-7
戻る
0
views
See details
Views

Versions

Ver.1 2023-09-20 08:09:38.934031
Show All versions

エクスポート

OAI-PMH
  • OAI-PMH JPCOAR
  • OAI-PMH DublinCore
  • OAI-PMH DDI
Other Formats
  • JSON
  • BIBTEX

Confirm


Powered by WEKO3

Change consent settings


Powered by WEKO3

Change consent settings