2026-05-14T19:59:16Z https://ipsj.ixsq.nii.ac.jp/oai

oai:ipsj.ixsq.nii.ac.jp:00017531 2025-01-22T23:16:44Z 934:1022:1044:1045

PO - P : A Concurrency Control Protocol for Parallel B - trees PO - P : A Concurrency Control Protocol for Parallel B - trees Damdinsuren, Amarmend Masayoshi, Aritsugi Yoshinari, Kanamori Damdinsuren, Amarmend Masayoshi, Aritsugi Yoshinari, Kanamori 研究論文 A number of parallel and distributed B-tree structures for shared-nothing parallel database systems have been proposed in connection with the related technological advancements such as clustered computers. We focus on the performance of the concurrency control protocols for such parallel B-trees and consider the possibility and the suitability of introduction of preparatory operations such as preemptive node split or merge to them in this paper. Since the early top-down concurrency control protocols were only designed for shared-everything architecture it is inappropriate to directly apply them to the distributed environment because the way update processes traverse the B-tree may cause significant degradations in the system performance. In this paper we propose a protocol named preparatory operations-parallel (PO-P) in which such preparatory operations are used during traversals of update processes. The results of our implementation of PO-P on a parallel B-tree in comparison with the existing protocol INC-OPT prove that the PO-P can serve as a good alternative protocol for distributed parallel B-trees. A number of parallel and distributed B-tree structures for shared-nothing parallel database systems have been proposed in connection with the related technological advancements such as clustered computers. We focus on the performance of the concurrency control protocols for such parallel B-trees and consider the possibility and the suitability of introduction of preparatory operations such as preemptive node split or merge to them in this paper. Since the early top-down concurrency control protocols were only designed for shared-everything architecture, it is inappropriate to directly apply them to the distributed environment because the way update processes traverse the B-tree may cause significant degradations in the system performance. In this paper, we propose a protocol, named preparatory operations-parallel (PO-P), in which such preparatory operations are used during traversals of update processes. The results of our implementation of PO-P on a parallel B-tree in comparison with the existing protocol INC-OPT prove that the PO-P can serve as a good alternative protocol for distributed parallel B-trees. journal article 情報処理学会 2004-12-15 application/pdf 情報処理学会論文誌データベース（TOD） SIG14(TOD24) 45 30 38 1882-7799 AA11464847 https://ipsj.ixsq.nii.ac.jp/record/17531/files/IPSJ-TOD4514005.pdf eng