Optimal Checkpointing Policies Using the Checkpointing Density

• Item type: JInfP(1)
• 公開日: 1992-03-31
• タイトル: Optimal Checkpointing Policies Using the Checkpointing Density
• タイトル / 言語: en
• タイトル: Optimal Checkpointing Policies Using the Checkpointing Density
• 言語: eng
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_6501
• 資源タイプ: journal article
• 著者所属: Graduate School of Engineering Hiroshima University
• 著者所属: Department of Management Science Hiroshima Shudo University
• 著者所属: Cluster II (Electrical Engineering) Faculty of Engineering Hiroshima University
• 著者所属(英): en
• 著者所属(英): Graduate School of Engineering, Hiroshima University
• 著者所属(英): en
• 著者所属(英): Department of Management Science, Hiroshima Shudo University
• 著者所属(英): en
• 著者所属(英): Cluster II (Electrical Engineering) Faculty of Engineering, Hiroshima University
• 著者名: Satoshi, Fukumoto; Naoto, Kaio; Shunji, Osaki
• 著者名(英): Satoshi, Fukumoto; Naoto, Kaio; Shunji, Osaki
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: In a computer system in particular in a database system constitutions and recovery techniques of fi1es play an important role in the operation of the system. This paper discusses checkpointing policies for rollback-recovery which is one of the most general file recovery techniques. When files on a main memory are lost in a failure we reprocess transactions from the latest checkpoint instead of the starting point of the system operation. Checkpoints are prespecified time points at which the information of the files is collected in a stable secondary storage. It is important to decide an effective checkpointing policy. If we execute checkpointing frequently we must incur a great cost for the collection of the information and conversely if we execute rare checkpointing we must incur a great cost for the recovery action after the system failure. Thus checkpoints should be decided by considering the trade-off between the above two costs. In this paper we discuss the checkpoint time sequence which minimizes the approximate total expected cost per unit time in the steady-state. The analysis shows the derivation of the total cost as the functional of the checkpointing density and the optimal checkpoints. We further present numerical examples assuming a failure-time distribution to be a Weibull distribution and show how our analytical results are of great use.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: In a computer system, in particular, in a database system, constitutions and recovery techniques of fi1es play an important role in the operation of the system. This paper discusses checkpointing policies for rollback-recovery which is one of the most general file recovery techniques. When files on a main memory are lost in a failure, we reprocess transactions from the latest checkpoint instead of the starting point of the system operation. Checkpoints are prespecified time points at which the information of the files is collected in a stable secondary storage. It is important to decide an effective checkpointing policy. If we execute checkpointing frequently, we must incur a great cost for the collection of the information, and conversely, if we execute rare checkpointing, we must incur a great cost for the recovery action after the system failure. Thus, checkpoints should be decided by considering the trade-off between the above two costs. In this paper, we discuss the checkpoint time sequence which minimizes the approximate total expected cost per unit time in the steady-state. The analysis shows the derivation of the total cost as the functional of the checkpointing density and the optimal checkpoints. We further present numerical examples assuming a failure-time distribution to be a Weibull distribution, and show how our analytical results are of great use.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AA00700121
• 書誌情報: Journal of Information Processing; 巻 15, 号 1, p. 97-92, 発行日 1992-03-31
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 1882-6652
• 出版者 / 言語: ja
• 出版者: 情報処理学会