System-R关系型数据库论文_写一篇关于数据库的英文论文

需积分: 0 529 浏览量更新于2023-05-27 评论收藏 3.03MB PDF 举报

身份认证购VIP最低享 7 折!

领优惠券(最高得80元）

资源详情

资源评论

资源推荐

System R: Relational Approach to Database

Management

M. M. ASTRAHAN, ht. W. BLASGEN, D. D. CHAMBERLIN,

K. P. ESWARAN, J. N. GRAY, P. P. GRIFFITHS,

W. F. KING, R. A. LORIE, P. R. A&JONES, J. W. MEHL,

G. R. PUTZOLU, I. L. TRAIGER, B. W. WADE, AND V. WATSON

IBM Research Laboratory

System R is a database management system which

provides a high level relational data interface.

The system provides a high level of data independence by isolating the end user as much as

possible from underlying storage structures. The system permits definition of a variety of relational

views on common underlying data. Data control features are provided, including authorization,

integrity assertions, triggered transactions, a logging and recovery subsystem, and facilities for

maintaining data consistency in a shared-update environment.

This paper contains a description of the overall architecture and design of the system. At the

present time the system is being implemented and the design evaluated. We emphasize that

System R is a vehicle for research in database architecture, and is not planned as a product.

Key Words and Phrases: database, relational model, nonprocedural language, authorization,

locking, recovery, data structures, index structures

CR categories: 3.74, 4.22, 4.33, 4.35

1. INTRODUCTION

The relational model of data was introduced by Codd [7] in 1970 as an approach

toward providing solutions to various problems in database management. In par-

ticular, Codd addressed the problems of providing a data model or view which is

divorced from various implementation considerations (the data independence

problem) and also the problem of providing the database user with a very high

level, nonprocedural data sublanguage for accessing data.

To a large extent, the acceptance and value of the relational approach hinges on

the demonstration that a system can be built which can be used in a real environ-

ment to solve real problems and has performance at least comparable to today’s

existing systems. The purpose of this paper is to describe the overall architecture

and design aspects of an experimental prototype database management system

called System R, which is currently being implemented and evaluated at the IBM

San Jose Research Laboratory. At the time of this writing, the design has been

but not for profit, all or part of this material is granted provided that ACM’s copyright notice is

given

and that reference

is made to the publication, to its date of issue, and to the fact that

reprinting privileges were granted by permission of the Association for Computing Machinery.

Authors’ address: IBM Research Laboratory, San Jose, CA 95193.

ACM Transactions on Database Systems, Vol. 1, No. 2. June 1976, Pages 97-137.

M. M. Astrahan et al,

CONTENTS

INTRODUCTION

Architecture and System Structure

2. THE RELATIONAL DATA SYSTEM

Host Language Interface

Query Facilities

Data Manipulation Facilities

Data Definition Facilities

Data Control Facilities

The Optimizer

Modifying Cursors

Simulation of Nonrelational Data Models

3. THE RELATIONAL STORAGE SYSTEM

Segments

Relations

Images

Links

Transaction Management

Concurrency Control

System Checkpoint and Restart

4. SUMMARY AND CONCLUSION

APPENDIX I. RDI Operators

APPENDIX II.

SEQUEL

Syntax

APPENDIX III. RSI Operators

ACKNOWLEDGMENTS

REFERENCES

completed and major portions of the system are implemented and running. How-

ever, the overall system is not completed. We plan a complete performance evalua-

tion of the system which will be available in later papers.

The System R project is not the first implementation of the relational approach

[12, 301. On the other hand, we know of no other relational system which provides

a complete database management capability-including application programming

as well as query capability, concurrent access support, system recovery, etc. Other

relational systems have focused on, and demonstrated, feasibility of techniques

for solving various specific problems. For example, the IS/l system [22] demon-

strated the feasibility of supporting the relational algebra [S] and also developed

optimization techniques for evaluating algebraic expressions [29]. Techniques for

optimization of the relational algebra have also been developed by Smith and

Chang at the University of Utah [27]. The extended relational memory (XRM)

system [19] developed at the IBM Cambridge Scientific Center has been used as

a single user access method by other relational systems [2]. The

SEQUEL

prototype

[l] was originally developed as a single-user system to demonstrate the feasibility

of supporting the

SEQUEL

[5] language. However, this system has been extended

by the IBM Cambridge Scientific Center and the MIT Sloan School Energy Labor-

atory to allow a simple type of concurrency and is being used as a component of

the Generalized Management Information System (GMIS) [9] being developed

at MIT for energy related applications. The INGRES project [lS] being developed

at the University of California, Berkeley, has demonstrated techniques for the de-

composition of relational expressions in the

QUEL

language into “one-variable

ACM Transactions on Database Systems, Vol. 1, No. 2, June 1976.

System R

queries.” Also, this system has investigated the use of query modification [28] for

enforcing integrity constraints and authorization constraints on users. The problem

of translating a high level user language into lower level access primitives has also

been studied at the University of Toronto [21,26].

Architecture and System Structure

We will describe the overall architecture of System R from two viewpoints. First,

we will describe the system as seen by a single transaction, i.e. a monolithic de-

scription. Second, we will investigate its multiuser dimensions. Figure 1 gives a

functional view of the system including its major interfaces and components.

The Relational Storage Interface (RSI) is an internal interface which handles

access to single tuples of base relations. This interface and its supporting system,

the Relational Storage System (RSS) , is actually a complete storage subsystem in

that it manages devices, space allocation, storage buffers, transaction consistency

and locking, deadlock detection, backout, transaction recovery, and system re-

covery. Furthermore, it maintains indexes on selected fields of base relations, and

pointer chains across relations.

The Relational Data Interface (RDI) is the external interface which can be

called directly from a programming language, or used to support various emulators

and other interfaces. The Relational Data System (RDS), which supports the

RDI, provides authorization, integrity enforcement, and support for alternative

views of data. The high level

SEQUEL

language is embedded within the RDI, and

is used as the basis for all data definition and manipulation. In addition, the RDS

maintains the catalogs of external names, since the RSS uses only system generated

internal names. The RDS contains an optimizer which chooses an appropriate

access path for any given request from among the paths supported by the RSS.

P-m

Programs to support

+ various Interfaces:

Stand alone SEQUEL,

Query By Example, etc.

- Relational

Data

RelatIonal

Interface

DC3

(RDII

System

IRDSI

RDS

RSS

I :

I ’ I

Monitor

storage

Interface

(RSll

FIQ. 1. Architecture of System R

FIG. 2. Use of virtual machines

in System R

ACM Transactions on Database Systems. Vol. 1, No. 2, June 1878.

100

M. M. Astrahan et al.

The current operating system environment for this experimental system is

VM/370 [lS]. Several extensions to this virtual machine facility have been made

[14] in order to support the multiuser environment of System R. In particular,

we have implemented a technique for the selective sharing of read/write virtual

memory across any number of virtual machines and for efficient communication

among virtual machines through processor interrupts. Figure 2 illustrates the use

of many virtual machines to support concurrent transactions on shared data. For

each logged-on user there is a dedicated

database machine.

Each of these database

machines contains all code and tables needed to execute all data management

functions; that is, services are not reserved to a centralized machine.

The provision for many database machines, each executing shared, reentrant

code and sharing control information, means that the database system need not

provide its own multitasking to handle concurrent transactions. Rather, one can

use the host operating system to multithread at the level of virtual machines.

Furthermore, the operating system can take advantage of multiprocessors allo-

cated to several virtual machines, since each machine is capable of providing all

data management services. A single-server approach would eliminate this advan-

tage, since most processing activity would then be focused on only one machine.

In addition to the database machines, Figure 2 also illustrates the Monitor

Machine, which contains many system administrator facilities. For example, the

Monitor Machine controls logon authorization and initializes the database machine

for each user. The Monitor also schedules periodic checkpoints and maintains

usage and performance statistics for reorganization and accounting purposes.

In Sections 2 and 3 we describe the main components of System R: the Relational

Data System and the Relational Storage System.

2. THE RELATIONAL DATA SYSTEM

The Relational Data Interface (RDI) is the principal external interface of System

R. It provides high level, data independent facilities for data retrieval, manipula-

tion, definition, and control. The data definition facilities of the RDI allow a variety

of alternative relational views to be defined on common underlying data. The

Relational Data System (RDS) is the subsystem which implements the RDI. The

RDS contains an optimizer which plans the execution of each RDI command,

choosing a low cost access path to data from among those provided by the Rela-

tional Storage System (RSS) .

The RDI consists of a set of operators which may be called from PL/I or other

host programming languages. (See Appendix I for a list of these operators.) All

the facilities of the

SEQUEL

data sublanguage [S] are available at the RDI by

means of the RDI operator called SEQUEL. (A Backus-Naur Form (BNF) syntax

for

SEQUEL

is given in Appendix II.) The

SEQUEL

language can be supported as a

stand-alone interface by a simple program, written on top of the RDI, which

handles terminal communications. (Such a stand-alone

SEQUEL

interface, called

the User-Friendly Interface, or UFI, is provided as a part of System R.) In addi-

tion, programs may be written on top of the RDI to support other relational inter-

faces, such as Query by Example [31], or to simulate nonrelational interfaces.

ACM TransactionsonDstabase Syystans,Vol. l.No. 2, June 1976.

System R

101

Host Language Interface

The facilities of the RDI are basically those of the SEQUEL data sublanguage, which

is described in [5] and in Appendix II. Several changes have been made to

SEQUEL

since the earlier publication of the language; they are described below.

The illustrative examples used in this section are based on the following database

of employees and their departments:

EMP(EMPN0, NAME, DNO, JOB, SAL, MGR)

DEPT(DN0, DNAME, LOC, NEMPS)

The RDI interfaces

SEQUEL

to a host programming language by means of a

concept called a cursor, A cursor is a name which is used at the RDI to identify a

set of tuples called its active

set

(e.g. the result of a query) and furthermore to main-

tain a position on one tuple of the set. The cursor is associated with a set of tuples

by means of the RDI operator SEQUEL; the tuples may then be retrieved, one at a

time, by the RDI operator FETCH.

Some host programs may know in advance exactly the degree and data types

of the tuples they wish to retrieve. Such a program may specify, in its SEQUEL call,

the program variables into which the resulting tuples are to be delivered. The pro-

gram must first give the system the addresses of the program variables to be used

by means of the RDI operator BIND, In the following example, the host program

identifies variables X and Y to the system and then issues a query whose results

are to be placed in these variables:

CALL BIND (‘Xl, ADDR(X)) ;

CALL BIND(‘Y’, ADDR(Y));

CALL SEQUEL(C1, ‘SELECT NAME:X, SAL:Y

FROM EMP

WHERE JOB = ’ ‘PROGRAMMER’ ’ ‘);

The SEQUEL call has the effect of associating the cursor Cl with the set of tuples

which satisfy the query and positioning it just before the first such tuple. The

optimizer is invoked to choose an access path whereby the tuples may be material-

ized. However, no tuples are actually materialized in response to the SEQUEL call.

The materialization of tuples is done as they are called for, one at a time, by the

FETCH operator. Each call to FETCH delivers the next tuple of the active set

into program variables X and Y, i.e. NAME to X and SAL to Y:

CALL FETCH(C1) ;

A program may wish to write a

SEQUEL

predicate based on the contents of a

program variable-for example, to find the programmers whose department num-

ber matches the contents of program variable 2. This facility is also provided by

the RDI BIND operator, as follows:

CALL BIND(‘X’, ADDR(X)) ;

CALL BIND(‘Y’, ADDR(Y));

CALL BIND(‘Z’, ADDR(2));

CALL SEQUEL(C1, ‘SELECT NAME:X, SAL:Y

FROM EMP

WHERE JOB = ’ ‘PROGRAMMER’ ’

AND DNO = 2’);

CALL FETCH(C1) ;

ACM Tramactions on Database System, Vol. 1, No. 2, June 1976.

剩余40页未读，继续阅读

qq_31967983

粉丝: 0
资源: 1

会员权益专享

System-R关系型数据库论文

评论0

会员权益专享

最新资源

System-R关系型数据库论文

评论0

关系型数据库和非关系型数据库

《PostGIS入门》---GIS数据库的学习与应用

关系型数据库与非关系型数据库的相同点

关系型数据库和非关系型数据库的区别

非关系型数据库和关系型数据库的区别

非关系型数据库同关系型数据库都需要规范化数据

关系型和非关系型数据库的区别

非关系型数据库研究现状

关系型数据库的研究现状

有什么关系型数据库的论文综述推荐吗？

关系型数据库和非关系型数据库最根本的区别在哪？

请说一下使用关系型数据库的做法

为什么非关系型数据库查找数据效率高？

采用实体联系模型(E-R)设计数据库的概念结构。E-R模型的组成部分是实体、属性和关联。E-R模型用E-R图表示，向用户提示相关的内容。属性描述了实体的特征。降重

hbase-与传统关系数据库的对比

关系型数据库各自优点缺点

阐述分布式数据库Hbase和传统的关系型数据库的主要区别。

数据库如何将E-R模型转换为关系模型

会员权益专享

最新资源