Time database review
Abstract: This article is an overview of the Timed Database, the summary introduction of the background, research trend, then introduces the classification of the Timed database and the existing comparison model, and finally introduce the timeliness management and timeliness Database design problems.
l background
The traditional database technology reflects the data in the real world, but it only reflects the current state of the data in the real world, only the status of an object is at a certain moment (snapshot), does not contact in the past and future. This is what people often say that snapshot database. The modern information stream contains Temport Information, where time information, time interval information, and relative time information (before, after, overlapping), and the like.
The increasingly wide database application requires historic information that manages the handled event, and turns information of the system in the system. It is necessary to solve two problems: First, the historic information of the processed event is required, such as historical information related to personnel, finance, financial and natural disasters, and can see the essential law of things in things; second, require management database Times information of the system in the system, such as the time and time interval of the inspection, deletion, and the time standard for locking queuing queuing and resource competition in multi-user systems, which help to improve the reliability and efficiency of the database system. . So introduce a temptation database.
l research trend
In the early 1970s, many people began to explore the processing time information in the database and information systems.
At the beginning of the 1980s, Times database research began to prosper, in 1982-1986, there were more than 80 papers about Timed Database. By the mid-1980s, nearly 100 turns of time databases have been proposed. After more than ten years of development and mutual learning, more than ten recognized models have gradually merged.
Kahn Ketal studied the issue of "Times" in 1977 in the article "Artificial Intelligence" magazine.
J. BEN ZVI has a pioneering study of Tetheet database in 1979-1982, his doctoral thesis (California University Los Angeles, 1982) summed up his series of work. Ben ZVI's contribution highlights the following:
(1) He proposed a timed database model that introduces time period, which is later renamed as a time interval.
(2) At 1979-1982, the Ben ZVI breakthrough in the hotspot, Ben ZVI breakthrough, proposed and studied Non-1NF TDB.
(3) Taking the time interval value, refreshing people think that the database field value can only be a number or string concept.
(4) Introduced the concept that is later referred to as dual, that is, use the effective time to represent the lifecycle of the managed object in the library, using the transaction time to represent the history of the database itself.
(5). The timed index structure is introduced.
1982 J. Clifford completed the University of New York, "a logical frame for the temporal semantics and nature language querying of histical database" and related groups of articles and related groups of articles, which made important contributions to the pioneering of historical database. It noted that the life cycle of managed objects was studied, and the technical details of the relationship, tuple, field value, introduced the historical relationship model, historical relationship algebra, and studied the projection and choice of historical database. , The special requirements and special laws of the connection have studied the compatibility of the historical relationship model and the traditional relational model. That is, when the interval is reduced to a point [now, the historical database is degraded into the traditional snapshot database, and the corresponding episodes The calculation degradation is a traditional snapshot relationship.
Professor S. Ginsburg, University of South California, is originally a formal language, especially a pioneer in the context of unreasonable research. He proposed an object history model (Object History) in 1983. l Times database classification
According to Spipada and Snodgrass, Timed Database can be divided into three categories:
(1) Transaction database. The database itself is called the transaction time, the transaction time, the transaction database supports the transaction time, and he puts all state evolution in the transaction time. The relationship between the transaction is a three-dimensional structure, consisting of a tuple, attribute, and transaction time.
(2) Historical database. The life cycle of managed objects is called the valid time, the historical database, and the transaction database are similar, just replacing the transaction time with the effective time, in which a historic state is recorded in each relationship. The relationship between history is also a three-dimensional structure, consisting of a tuple, attribute, and effective time.
(3) Time database, can manage the history of objects, but also manage the history of the database itself, and call the dual-Time database. He has the advantages of the top two, supporting transaction time and effective time. Time relationship is a four-dimensional structure, consisting of a tuple, attribute, transaction time, and effective time.
L DBMS Times Information Four Processing Method
Events are often accompanied by time information, and traditional databases record temptation information only by user-defined time, and DBMS does not manage event native information. The following table lists four ways of DBMS handling temptation information.
l Times management problem
For a long time, in the age of not using the Timed Database, the management of the management of enterprises and organs is put together with multiple snapshots (ie, the backend of the database saved at different times), constitutes history.
(1). How much time interval saves a snapshot? If the interval is too large, it is not enough to ensure accurate data from the data. If the interval is too small, the data is redundant, and the storage space is large.
(2). In the traditional relational database, multiple snapshots of a table cannot be simultaneously loaded into memory, and the traditional selection, projection, connection operation cannot be simultaneously used. Because the values under the same property of the same level may be different in different snapshots, it must be complex, non-universal programming.
(3). In the traditional database, the maintenance support for the database itself is insufficient, and generally only the transaction log is only available to return to the returned transaction log, lacking the corresponding transaction query command (such as quick findings who have made more than ten times for a field. Modifications).
l Existing Timed Database Model
In the middle of the 1980s, the program of nearly 100 turns has been proposed. After more than ten years of development and mutual learning, it gradually merged into more than ten recognized models, including 13 of the world's first world about the temporal Database monograph "Temport Database-theory, Design and Implementation".
(1). Time Relational Model, this model is Ben ZVI proposed in doctoral thesis in 1979--1982 and is a pioneering research in TDB. It created research on Timethical Database, Times Query.
(2). HRDM (Historical Relational Data Model), J. Clifford, 1982.
(3). Tempsql, Sharshi. K. Gadia & Sunil,
S. Nair
1985.
(4). IXRM (Interval-Extended Relational Model), Nikos A. Lorentzos, 1987.
(5). TRM and TSQL (Temporal Extensions to the Relational Model) K. B. Navathe, 1987. (6). HSQL (Historical Query Language), N. L. SARDA 1987.
(7). TQuel, R. Snodgrass, 1985.
(8). Trc (Temporal Relational Calculus), ABDULLAH TANSEL, 1992
(9). Teer, (Temporal Query Language for Enhanced Entity Relationship Model), R. Elmasri, 1985.
(10). TDM (Temporal Data Model Based on Time Sequence), Arie Segev & Arie Shosham, 1988.
(11). Oodaplex (Object Oriented APLEX), U. DAYAL, 1989.
(12). Object History,
S. Ginsburg
, Tanaka, Tang Changjie, 1983.
(13). Temporal Deductive Databases, Marianne Bandinet, etc. 1989.
The above 13 TDB models have established a set of terms, concepts, mathematical models independently, and form a separate theoretical system.
l Time database is difficult during the specific implementation process
(1) Data amount is large. Time databases are intended to manage historical data. Over time, new data source is constantly entering the database, in order to ensure that TDB is time-space efficiency of TDB, there must be an efficient data storage organization and Timel Index mechanism.
(2) In the actual application, the selection, projection and connection operations account for the main resources. Optimized to time selection, timing projections, and timing connections become the technical focus of TDB query optimization.
(3) On the index, due to the joining timed effect, the traditional HASH and B, B Tree need to expand in the tense to meet the requirements of TDBMS.
l Time data management and artificial intelligence
In the field of artificial intelligence, the main research mechanism is mainly studied. There are two types:
(1) Explanation of the main data and its connection This system can be used to support natural language understanding of time problems. J. The "Time Zone Calculation" is representative.
(2) Establishing a future activity plan This system can be used in the use of factory dispatching resources, called "intelligent scheduling", which is: In a certain time limit, it is formulated that the correct use of resources has resolved a given scheduled problem. plan.
l reference
"Special Database Technology" He Xingui, Tang Changjie Science Publishing House 2000
"Timed Database Technology" Huang Nan, Liu Aiqin Microcomputer Development 2002 No. 1
