Item type |
Symposium(1) |
公開日 |
2016-11-21 |
タイトル |
|
|
タイトル |
Long Distance Synchronization Method for In-Memory State Management |
タイトル |
|
|
言語 |
en |
|
タイトル |
Long Distance Synchronization Method for In-Memory State Management |
言語 |
|
|
言語 |
eng |
キーワード |
|
|
主題Scheme |
Other |
|
主題 |
ストレージ・記憶階層 |
資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_5794 |
|
資源タイプ |
conference paper |
著者所属 |
|
|
|
Hitachi, Ltd., Research & Development Group, Center for Technology Innovation-Information and Telecommunications |
著者所属 |
|
|
|
Hitachi, Ltd., Research & Development Group, Center for Technology Innovation-Information and Telecommunications |
著者所属(英) |
|
|
|
en |
|
|
Hitachi, Ltd., Research & Development Group, Center for Technology Innovation-Information and Telecommunications |
著者所属(英) |
|
|
|
en |
|
|
Hitachi, Ltd., Research & Development Group, Center for Technology Innovation-Information and Telecommunications |
著者名 |
Arif, Herusetyo Wicaksono
Kazuhide, Aikoh
|
著者名(英) |
Arif, Herusetyo Wicaksono
Kazuhide, Aikoh
|
論文抄録 |
|
|
内容記述タイプ |
Other |
|
内容記述 |
With the needs of protecting data consistency between multiple data centers for state management in transactional system, a long distance data synchronization technology is required, both for improving performance by distributed processing and for providing fail-over in the case of large-scale disaster. However, in existing method, maintaining data consistency causes network latency degradation for ensuring completion of synchronization, while maintaining network latency performance causes large number of data inconsistencies in the case of large-scale disaster, both of which are unacceptable. Therefore, we propose a synchronization method with two main characteristics: prioritization of transmission of identifier information along with transaction ordering information for faster processing, and confirmation of replication status prior to I/O operation. By applying the proposal to an in-memory KVS, we verified that the number of data inconsistencies caused by disaster can be reduced by 39% for 100 kB value size, without affecting the network latency performance. Finally, it can be estimated that the minimizing data inconsistencies while maintaining network latency performance can be achieved with our proposed long-distance synchronization method. |
論文抄録(英) |
|
|
内容記述タイプ |
Other |
|
内容記述 |
With the needs of protecting data consistency between multiple data centers for state management in transactional system, a long distance data synchronization technology is required, both for improving performance by distributed processing and for providing fail-over in the case of large-scale disaster. However, in existing method, maintaining data consistency causes network latency degradation for ensuring completion of synchronization, while maintaining network latency performance causes large number of data inconsistencies in the case of large-scale disaster, both of which are unacceptable. Therefore, we propose a synchronization method with two main characteristics: prioritization of transmission of identifier information along with transaction ordering information for faster processing, and confirmation of replication status prior to I/O operation. By applying the proposal to an in-memory KVS, we verified that the number of data inconsistencies caused by disaster can be reduced by 39% for 100 kB value size, without affecting the network latency performance. Finally, it can be estimated that the minimizing data inconsistencies while maintaining network latency performance can be achieved with our proposed long-distance synchronization method. |
書誌情報 |
コンピュータシステム・シンポジウム論文集
巻 2016,
p. 35-41,
発行日 2016-11-21
|
出版者 |
|
|
言語 |
ja |
|
出版者 |
情報処理学会 |