| Item type |
SIG Technical Reports(1) |
| 公開日 |
2017-03-01 |
| タイトル |
|
|
タイトル |
Improving Multi-component Application Performance by Silently Replacing File I/O with Direct Data Transfer: Preliminary Results |
| タイトル |
|
|
言語 |
en |
|
タイトル |
Improving Multi-component Application Performance by Silently Replacing File I/O with Direct Data Transfer: Preliminary Results |
| 言語 |
|
|
言語 |
eng |
| キーワード |
|
|
主題Scheme |
Other |
|
主題 |
タスクスケジューリング |
| 資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_18gh |
|
資源タイプ |
technical report |
| 著者所属 |
|
|
|
RIKEN AICS |
| 著者所属 |
|
|
|
Northwestern University |
| 著者所属 |
|
|
|
RIKEN AICS |
| 著者所属 |
|
|
|
RIKEN AICS |
| 著者所属 |
|
|
|
Northwestern University |
| 著者所属(英) |
|
|
|
en |
|
|
RIKEN AICS |
| 著者所属(英) |
|
|
|
en |
|
|
Northwestern University |
| 著者所属(英) |
|
|
|
en |
|
|
RIKEN AICS |
| 著者所属(英) |
|
|
|
en |
|
|
RIKEN AICS |
| 著者所属(英) |
|
|
|
en |
|
|
Northwestern University |
| 著者名 |
Tatiana, V. Martsinkevich
Wei-Keng, Liao
Balazs, Gerofi
Yutaka, Ishikawa
Alok, Choudhary
|
| 著者名(英) |
Tatiana, V. Martsinkevich
Wei-Keng, Liao
Balazs, Gerofi
Yutaka, Ishikawa
Alok, Choudhary
|
| 論文抄録 |
|
|
内容記述タイプ |
Other |
|
内容記述 |
With the growing performance of current HPC systems, we can afford applying more sophisticated algorithms to processing the ever increasing volumes of data. This includes the multi-component execution scenario where each component performs a particular data transformation, then outputs the result to a file, which in turn, becomes the input for the next component. However, File I/O is known to impede scalability. To improve the execution performance we developed a framework that replaces File I/O with direct data transfer. Using the framework requires minimal modification of the application and the data transfer happens transparently for the user. While the current implementation of the framework specifically targets applications that use PnetCDF for File I/O, we note that the framework is general and can be adapted to use any other I/O library. |
| 論文抄録(英) |
|
|
内容記述タイプ |
Other |
|
内容記述 |
With the growing performance of current HPC systems, we can afford applying more sophisticated algorithms to processing the ever increasing volumes of data. This includes the multi-component execution scenario where each component performs a particular data transformation, then outputs the result to a file, which in turn, becomes the input for the next component. However, File I/O is known to impede scalability. To improve the execution performance we developed a framework that replaces File I/O with direct data transfer. Using the framework requires minimal modification of the application and the data transfer happens transparently for the user. While the current implementation of the framework specifically targets applications that use PnetCDF for File I/O, we note that the framework is general and can be adapted to use any other I/O library. |
| 書誌レコードID |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AN10463942 |
| 書誌情報 |
研究報告ハイパフォーマンスコンピューティング（HPC）
巻 2017-HPC-158,
号 33,
p. 1-5,
発行日 2017-03-01
|
| ISSN |
|
|
収録物識別子タイプ |
ISSN |
|
収録物識別子 |
2188-8841 |
| Notice |
|
|
|
SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc. |
| 出版者 |
|
|
言語 |
ja |
|
出版者 |
情報処理学会 |
IPSJ-HPC17158033.pdf (7.4 MB)
