Efficient time-interval data extraction in MVCC-based RDBMS
Efficient time-interval data extraction in MVCC-based RDBMS
复制标题
DOI：
10.1007/s11280-018-0552-7
复制
发表时间：
2018-04
期刊：
影响因子：
--
通讯作者：
Haixiang Li;Zhanhao Zhao;Yijian Cheng;Wei Lu;Xiaoyong Du;Anqun Pan
中科院分区：
文献类型：
--
作者：
Haixiang Li;Zhanhao Zhao;Yijian Cheng;Wei Lu;Xiaoyong Du;Anqun Pan
对账是银行和游戏公司的核心业务。它定期检查每个用户的银行或费用报表的帐户余额，并报告每日，每周或每月的余额。一旦发生账户不平衡，有必要有效地跟踪可能破坏账户余额的交易。为了帮助有效地跟踪此类事务，本文研究了在基于MVCC的关系数据库管理系统中进行有效的时间间隔数据提取的问题，即，提取在基于MVCC的RDBMS中的给定时间间隔之间有效的增量数据。为此，我们提出了一个基于快照的方法来提取增量数据的基础上，每个记录是固有的与生命周期，指示是否可以访问记录或不为给定的时间间隔。我们阐述了如何将我们的方法集成到MySQL，一个开源的RDBMS，并提出了一个声明式的方式来获取增量数据。提出了几种优化技术来提高提取性能。在标准化的Sysbench基准上进行了大量的实验，结果表明，我们提出的方法是鲁棒和有效的。
Account reconciliation is the core business in banks and game companies. It regularly examines the account balance with the bank or expense statement for every user and reports the daily, weekly, or monthly balance. Once an account imbalance occurs, it is necessary to efficiently trace the transactions that possibly destroy the account balances. To help efficiently trace this kind of transactions, in this paper, we investigate the problem of doing efficient time-interval data extraction in MVCC-based RDBMS, i.e., extracting the incremental data that are valid between a given time interval in MVCC-based RDBMS. To this end, we propose a snapshot-based method to extract incremental data based on the fact that each record is inherently associated with lifetime, indicating whether the record can be accessed or not for a given time interval. We elaborate how to integrate our method into MySQL, an open-sourced RDBMS, and propose a declarative way to fetch the incremental data. Several optimization techniques are proposed to boost the extraction performance. Extensive experiments are conducted over the standardized Sysbench benchmark to show that our proposed method is robust and efficient.