Tcrnporal Managcment of DataBase
There are two components to temporal datamanagement: history data management and versionmanagement. Both have been the subjects of researchfor over a decade. The troublesome aspect of tempo-ral data management is that the boundary betweenapplications and database systems has not beenclearly drawn.Specifically, it is not clear how muchof the typical semantics and facilities of temporaldata management can and should be directly incorpo-rated in a database systqm,and how much should belcft to applications and users.In this section, we willprovide a list of short-term research issues thatshould be examined to shed light on this fundamentalquestion.
The focus of research into history data manage-ment has been on defining the semantics of time andtime interval, and issues related to understanding thesemantics of queries and updates against histo叮datastored in an attribute of a record. Typically, in thecontext of relational databases,a temporal attribute isdefined to hold a sequence of history data for theattribute. A history data consists of a data item and atime interval for which the data item is valid. A quenmav- then be issued to retrieve historv data for aspecified time intei-val for the temporal attribute. Themechanism for supporting temporal attributes is simi-lar to that for supporting set-valued attributes in adatabase system, such as UnISQL
In the absence of a support for temporal attrib-utes, application developers who need to model andmanage history data have simply simulated temporalattributes by creating attributc(s) for the time inter-val, along with the" temporal" attribute. This ofcourse may result in duplication of records in a table,and more complicated search predicates in queries.The one necessary topic of research in history data.nanagement is to quantitatively establish the per-.ormance (and even productivity) differences betweenasing a database system that directly supports attributes and using a conventional DataBascs,system that does not support either the set-valuedattributes or temporal attributes.If it can be convinc ngly established that the benefits of a DataBasc, in that supports temporal attributes arc substantial,DataBase vendors will be strongly motivated to aug-ment their systems with history data management.
翻译:
数据库的时间组织
时间数据管理有两个组成部分:历史数据管理和文本管理。这两种管理多年来一直是研究的对象。时间数据管理方面的麻烦是: 应用和数据库系统之间的界限还没有划清楚。特别是时间数据管理的典型语义和设备有多少能够月应该直接合并在数据库系统中,有多少应该留给应用和用户,还不清楚。 在这一章中,我们将提供一个短期研究问题目录,应该检查该目录以阐明这一基本问题。
研究历史数据库的焦点是确定时间和时间间隔的语义和涉及了解 询问和更改存储在记录特征的历史数据库语义问题。典型地说,在关系数据库中,时间属性被确定为持有顺序的具有属性的历史数据, 一个历史数据是由数据项目和对数据项目的有效的时间间隔组成。在给定的时间间隔,就会产生一次查询,即为该特定时间间隔而检索历史数据。支持时间属性的作用相似于支持数据库系统中的定值属性。如UnISQL。
在时间属性没有支持的情况下,需要模拟和管理历史数据,发展者们通过为时间间隔及时一间属性创造属性而开始简单地模拟属性。这当然可能会在表格中等致记录的复制及询问更加复杂的搜索依据。在有关历史研究的数据管理中,一个必要的研究课题是定量地建立效率差别,通过使用直接支持时间属性的数据库系统和使用不支持时间属性的常规数据库系统。如果能够很有说服力地建立起这样一种观点:一个支持时间属性的数据库系统会带来很大的利益,那么会促使数销售商极大地用历史数据管理来扩大它们的系统。
责任编辑:cyth