Object-Relational DBMSObject-Relational DBMSss

By Yao-Wen TuBy Yao-Wen Tu

CS157bCS157b12/09/200312/09/2003

Prof. Sin-Min LeeProf. Sin-Min Lee

Introduction to Object-Relational DIntroduction to Object-Relational DBMSsBMSs

Several major software companies incluSeveral major software companies including ding IBMIBM, , InformixInformix, , MicrosoftMicrosoft, , OracleOracle, an, and d SybaseSybase have all released object-relatio have all released object-relational versions of their products. These conal versions of their products. These companies are promoting a new, extended mpanies are promoting a new, extended version of relational database technologversion of relational database technology called y called object-relational database maobject-relational database management systemsnagement systems also known as also known as ORDORDBMSsBMSs. .

Introduction to Object-Relational DIntroduction to Object-Relational DBMSs (Cont)BMSs (Cont)

This article compares and contrasts this This article compares and contrasts this new class of database with the relationanew class of database with the relational databases, RDBMS from which they are l databases, RDBMS from which they are evolving and also with efficient object-orevolving and also with efficient object-oriented databases, OODBMSs, also knowiented databases, OODBMSs, also known as object databases, ODBMSs. n as object databases, ODBMSs.

Does a database supporting complex applicDoes a database supporting complex applic

ations have to be object-oriented?ations have to be object-oriented?

A certain group thinks that future applications can onlA certain group thinks that future applications can only be implemented with pure object-oriented systems. y be implemented with pure object-oriented systems. Initially these systems looked promising. However, thInitially these systems looked promising. However, they have been unable to live up to the expectations. A ney have been unable to live up to the expectations. A new technology has evolved in which relational and objew technology has evolved in which relational and object-oriented concepts have been combined or mergeect-oriented concepts have been combined or merged. These systems are called d. These systems are called object-relational database object-relational database systemssystems. The main advantages of ORDBMSs are massi. The main advantages of ORDBMSs are massive scalability and support for object-oriented features.ve scalability and support for object-oriented features.

Advantages of ORDBMSsAdvantages of ORDBMSs

The main advantages of extending the The main advantages of extending the relational data model come from relational data model come from reusereuse and and sharingsharing..

Reuse comes from the ability to extend the Reuse comes from the ability to extend the DBMS server to perform standard DBMS server to perform standard functionality centrally, rather than have it functionality centrally, rather than have it coded in each application.coded in each application.

If we can embed the functionality in the If we can embed the functionality in the server, it saves having to define it in each server, it saves having to define it in each application that needs it, and consequently application that needs it, and consequently allows the functionality to be shared by all allows the functionality to be shared by all applications.applications.

Disadvantages of ORDBMSsDisadvantages of ORDBMSs The ORDBMSs approach has the obvious disadvantagThe ORDBMSs approach has the obvious disadvantag

e of complexity and associated increased costs.e of complexity and associated increased costs.

There are proponents of the relational approach that There are proponents of the relational approach that believe the essential simplicity and purity of the relatibelieve the essential simplicity and purity of the relational model are lost with these types of extension.onal model are lost with these types of extension.

There are also those that believe that the RDMSs is beThere are also those that believe that the RDMSs is being extended for what will be a minority of applicationing extended for what will be a minority of applications that do not achieve optimal performance with curres that do not achieve optimal performance with current relational technology.nt relational technology.

Disadvantages of ORDBMSs (ConDisadvantages of ORDBMSs (Cont.)t.)

Instead of discussing object models, terms like ‘useInstead of discussing object models, terms like ‘user-defined data types’ are used. The terminology of or-defined data types’ are used. The terminology of object-orientation abounds with terms like ‘abstract tbject-orientation abounds with terms like ‘abstract types’. ‘class hierarchies’, and ‘object models’. ypes’. ‘class hierarchies’, and ‘object models’. However, ORDBMSs vendors are attempting to portraHowever, ORDBMSs vendors are attempting to portray object models as extensions to the relational model y object models as extensions to the relational model with some additional complexities.with some additional complexities.

This potentially misses the point of object-orientation,This potentially misses the point of object-orientation, highlighting the large semantic gap between these tw highlighting the large semantic gap between these two technologies.o technologies.

Relational DBMSRelational DBMS

The relational model was formally introduced The relational model was formally introduced by Dr. E. F. Codd in 1970 and has evolved since by Dr. E. F. Codd in 1970 and has evolved since then, through a series of writings and later thrthen, through a series of writings and later through implementations by IBM and others. The ough implementations by IBM and others. The defining standard for relational databases is pdefining standard for relational databases is published by ANSI (the American National Stanublished by ANSI (the American National Standard Institute) as SQL (ANSI 1986) or SQL1, calldard Institute) as SQL (ANSI 1986) or SQL1, called SQL-86. A revised standard is called SQL2, aed SQL-86. A revised standard is called SQL2, also referred to as SQL-92. lso referred to as SQL-92.

Relational DBMSRelational DBMS A relational database is composed of many relA relational database is composed of many rel

ations in the form of two-dimensional tables oations in the form of two-dimensional tables of rows and columns containing related tuples. f rows and columns containing related tuples.

Organizing data into tables, the form in which Organizing data into tables, the form in which data is presented to the user and the programdata is presented to the user and the programmer, is known as the mer, is known as the logical viewlogical view of the datab of the database. The stored data on a computer disk systease. The stored data on a computer disk system is called the m is called the internal viewinternal view. The rows (tuple. The rows (tuples) are called s) are called recordsrecords and the columns (fields i and the columns (fields in the record) are called n the record) are called attributesattributes. .

Relational DBMSRelational DBMS

Each column has a data type (i.e., int, floEach column has a data type (i.e., int, float, date). There are various restrictions oat, date). There are various restrictions on the data that can be stored in a relation the data that can be stored in a relational database. These are called nal database. These are called constraiconstraintsnts. The constraints are domain constrai. The constraints are domain constraints, key constraints, entity integrity consnts, key constraints, entity integrity constraints, and referential integrity constraitraints, and referential integrity constraints. These constraints ensure that there nts. These constraints ensure that there are no ambiguous tuples in the databasare no ambiguous tuples in the database. e.

RDBMSs RDBMSs

RDBMSs use Structured Query Language (SQL, RDBMSs use Structured Query Language (SQL, currently SQL2) as the data definition languagcurrently SQL2) as the data definition language (DDL) and the data manipulation language e (DDL) and the data manipulation language (DML). (DML).

SQL includes statements for data definition, mSQL includes statements for data definition, modification, querying and constraint specificatiodification, querying and constraint specification. The types of queries vary from simple singlon. The types of queries vary from simple single-table queries to complicated multi-table quee-table queries to complicated multi-table queries involving joins, nesting, set union/differenries involving joins, nesting, set union/differences, and others. ces, and others.

disadvantages of Relational Databadisadvantages of Relational Databasesses

All processing is based on values in fields of reAll processing is based on values in fields of records. Examples of RDBMSs include Oracle, decords. Examples of RDBMSs include Oracle, developed by Oracle Corporation, and Microsoft veloped by Oracle Corporation, and Microsoft Access developed by Microsoft. Access developed by Microsoft.

The main disadvantages of Relational DatabasThe main disadvantages of Relational Databases include their inability to handle application es include their inability to handle application areas like spatial databases (e.g. CAD), applicaareas like spatial databases (e.g. CAD), applications involving images, special types databasetions involving images, special types databases (e.g. complex numbers, arrays, etc.) and othes (e.g. complex numbers, arrays, etc.) and other applications that involve complex interrelatir applications that involve complex interrelationships of data. onships of data.

SQLSQL

The SQL standard enables users to easilThe SQL standard enables users to easily migrate their database applications bey migrate their database applications between database systems. tween database systems.

In addition, users can access data stored In addition, users can access data stored in two or more RDBMSs without changinin two or more RDBMSs without changing the database sub-language (SQL). g the database sub-language (SQL).

Object-Oriented DBMS (OODBMObject-Oriented DBMS (OODBMS)S)

The concept of The concept of abstract data types (ADTs)abstract data types (ADTs) in in which the internal data structure is hidden anwhich the internal data structure is hidden and the external operations can be applied on thd the external operations can be applied on the object that is specified led to the concept of e object that is specified led to the concept of encapsulation. encapsulation.

The programming language SMALLTALK, develThe programming language SMALLTALK, developed by Xerox, was designed to be object-orieoped by Xerox, was designed to be object-oriented. Other object-oriented programming langnted. Other object-oriented programming languages include C++, Java, etc. uages include C++, Java, etc.

Object-Oriented DBMS (OODBMObject-Oriented DBMS (OODBMS)S)

The main features of OO programming lThe main features of OO programming languages are encapsulation, inheritance anguages are encapsulation, inheritance and polymorphism. and polymorphism.

EncapsulationEncapsulation can be thought as a prot can be thought as a protective layer that prevents the code and tective layer that prevents the code and the data from being accessed by other cohe data from being accessed by other code defined outside the layer. de defined outside the layer.

Object-Oriented DBMS (OODBMObject-Oriented DBMS (OODBMS)S)

The process in which one object inherits the prThe process in which one object inherits the properties of a previously defined object is calleoperties of a previously defined object is called d inheritanceinheritance. Inheritance aids in the reuse of . Inheritance aids in the reuse of existing definitions for creating new objects. existing definitions for creating new objects.

PolymorphismPolymorphism allows the same operator or s allows the same operator or symbol to have different implementations, depymbol to have different implementations, depending on the type of objects to which the opeending on the type of objects to which the operator is applied. rator is applied.

object-oriented database managemobject-oriented database management systems (OODBMSs)ent systems (OODBMSs)

The data in object-oriented database mThe data in object-oriented database management systems (OODBMSs) is manaanagement systems (OODBMSs) is managed through two sets of relations, one dged through two sets of relations, one describing the interrelations of data items escribing the interrelations of data items and another describing the abstract relaand another describing the abstract relationships (inheritance). tionships (inheritance).

object-oriented database managemobject-oriented database management systems (OODBMSs)ent systems (OODBMSs)

The strong connection between applicatThe strong connection between application and database results in less code, mion and database results in less code, more natural data structures, and better ore natural data structures, and better maintainability and reusability of code. maintainability and reusability of code. OO languages, such as C++ or Java, are aOO languages, such as C++ or Java, are able to reduce code size by not having to ble to reduce code size by not having to translate code into a database sublangutranslate code into a database sublanguage such as SQL and ODBC or JDBC. age such as SQL and ODBC or JDBC.

Changing the lack of definingChanging the lack of defining

the lack of a defining standard was a drathe lack of a defining standard was a drawback for OODBMSs. The Object Data Mwback for OODBMSs. The Object Data Management Group (ODMG) has proposed anagement Group (ODMG) has proposed a standard known as ODMG-93 or ODMG a standard known as ODMG-93 or ODMG 1.0 standard , now revised into ODMG 2.1.0 standard , now revised into ODMG 2.0. 0.

ConsistsConsists

The standard consists of the object modThe standard consists of the object model, the object defining language (ODL), tel, the object defining language (ODL), the object query language (OQL), and the he object query language (OQL), and the bindings to OO programming languages. bindings to OO programming languages.

The ODL and OQL are based on the ODMThe ODL and OQL are based on the ODMG data model. G data model.

Data ModelData Model

The data model consists of data types, type coThe data model consists of data types, type constructors, etc., and is similar to the SQL repornstructors, etc., and is similar to the SQL report that describes the standard model for relatiot that describes the standard model for relational databases. nal databases.

The ODL is designed so as to support semantiThe ODL is designed so as to support semantic constructs of ODMG 2.0 object model. It is inc constructs of ODMG 2.0 object model. It is independent of any programming language. The dependent of any programming language. The ODL is used to create object specifications. ODL is used to create object specifications.

Data Model (cont.)Data Model (cont.)

The OQL is designed to work closely with the pThe OQL is designed to work closely with the programming languages for which an ODMG binrogramming languages for which an ODMG binding is defined such as C++, Java and SMALLTAding is defined such as C++, Java and SMALLTALK. LK.

The syntax of the OQL queries is similar to the The syntax of the OQL queries is similar to the syntax of SQL (a query language for relational syntax of SQL (a query language for relational databases) with some additional features such databases) with some additional features such as object identity, complex objects, inheritancas object identity, complex objects, inheritance, polymorphism and relationships. e, polymorphism and relationships.

OODBMSsOODBMSs

An object-oriented language is the languAn object-oriented language is the language for both the application and the datage for both the application and the database. OODBMSs have been integrated abase. OODBMSs have been integrated with C++, C, Java and LISP. with C++, C, Java and LISP.

The primary interface in an OODBMS for The primary interface in an OODBMS for creating and modifying objects is directlcreating and modifying objects is directly via the object language (C++, Java, ety via the object language (C++, Java, etc.) using the native language syntax. c.) using the native language syntax.

Difference between relational databDifference between relational databases and OO databasesases and OO databases

The difference between relational databases aThe difference between relational databases and OO databases is the way in which relationsnd OO databases is the way in which relationships are handled. hips are handled.

In OO databases, the relationships are represeIn OO databases, the relationships are represented with OIDs, which improves the data accesnted with OIDs, which improves the data access performance. s performance.

In relational databases, relationships among tIn relational databases, relationships among tuples are specified by attributes having the sauples are specified by attributes having the same domain. me domain.

OODBMSs ProblemsOODBMSs Problems

The main drawback of OODBMSs has been poThe main drawback of OODBMSs has been poor performance. Unlike RDBMSs, query optimior performance. Unlike RDBMSs, query optimization for OODBMs is highly complex. zation for OODBMs is highly complex.

OODBMSs also suffer from problems of scalabiOODBMSs also suffer from problems of scalability, and are unable to support large-scale systlity, and are unable to support large-scale systems. Some examples of OODBMSs are O2 (noems. Some examples of OODBMSs are O2 (now called Ardent) developed by Ardent Softwarw called Ardent) developed by Ardent Software, and the ObjectStore system produced by Obe, and the ObjectStore system produced by Object Design Inc. ject Design Inc.

Object-Relational DBMS (ORDBMObject-Relational DBMS (ORDBMS)S)

An ORDBMS supports an extended form of SQL An ORDBMS supports an extended form of SQL called SQL3 that is still in the development stacalled SQL3 that is still in the development stages. The extensions are needed because ORDBges. The extensions are needed because ORDBMSs have to support ADT's. MSs have to support ADT's.

The ORDBMS has the relational model in it becThe ORDBMS has the relational model in it because the data is stored in the form of tables haause the data is stored in the form of tables having rows and columns and SQL is used as the ving rows and columns and SQL is used as the query language and the result of a query is alsquery language and the result of a query is also table or tuples (rows). o table or tuples (rows).

characteristics of an ORDBMSs characteristics of an ORDBMSs

Base datatype extension, Base datatype extension, Support complex objects, Support complex objects, Inheritance, and Inheritance, and Rule Systems. Rule Systems.

Users define datatypes Users define datatypes

Object-Relational Database ManagermeObject-Relational Database Managerment Systems (ORDBMSs) allow users to dent Systems (ORDBMSs) allow users to define datatypes, functions and operators. fine datatypes, functions and operators. As a result, the functionality of the ORDBAs a result, the functionality of the ORDBMSs increases along with their performaMSs increases along with their performance. nce.

An example schema of a student reAn example schema of a student relation which ORDBMS supports lation which ORDBMS supports

STUDENT(fname,lname,ID,sex,major,address,STUDENT(fname,lname,ID,sex,major,address,dname,location,picture) dname,location,picture)

Notice : extra attributes "location" and "picturNotice : extra attributes "location" and "picture" which are not present in the traditional EMPe" which are not present in the traditional EMPLOYEE relation of RDBMS. The datatype of "locLOYEE relation of RDBMS. The datatype of "location" is "geographic point" and "picture" is "iation" is "geographic point" and "picture" is "image". mage".

The differences between the threThe differences between the three approachese approaches

CriteriaCriteria RDBMSRDBMS ODBMSODBMS ORDBMSORDBMS Defining standardDefining standard SQL2SQL2 ODMG-2.0ODMG-2.0 SQL3 (in process)SQL3 (in process)

Support for object-oriSupport for object-oriented featuresented features

Does not support; Does not support;

It is difficult to map prIt is difficult to map program object to the daogram object to the databasetabase

Supports extensivelySupports extensively Limited support; mostLimited support; mostly to new data typely to new data type

UsageUsage Easy to useEasy to use OK for programmers; OK for programmers; some SQL access for some SQL access for end usersend users

Easy to use except for Easy to use except for some extensionssome extensions

Support for complex rSupport for complex relationshipselationships

Does not support abstrDoes not support abstract datatypesact datatypes

Supports a wide varietSupports a wide variety of datatypes and daty of datatypes and data with complex inter-ra with complex inter-relationshipselationships

Supports Abstract dataSupports Abstract datatypes and complex reltypes and complex relationshipsationships

PerformancePerformance Very good performancVery good performancee

Relatively less perforRelatively less performancmanc

Expected to perform vExpected to perform very wellery well

The differences between the threThe differences between the three approachese approaches

CriteriaCriteria RDBMSRDBMS ODBMSODBMS ORDBMSORDBMS

Product maturity Product maturity Relatively old and so very Relatively old and so very mature mature

This concept is few years This concept is few years old and so relatively matuold and so relatively matur feature r feature

Still in development stage Still in development stage so immature so immature

The use of SQL The use of SQL Extensive supports SQL Extensive supports SQL OQL is similar to SQL, but OQL is similar to SQL, but with additional features liwith additional features like Complex objects and oke Complex objects and object-oriented features bject-oriented features

SQL3 is being developed SQL3 is being developed with OO features incorporwith OO features incorporated in it ated in it

AdvantagesAdvantages Its dependence on SQL, reIts dependence on SQL, relatively simple query optilatively simple query optimization hence good perfmization hence good performance ormance

It can handle all types of cIt can handle all types of complex applications, reusomplex applications, reusability of code, less coding ability of code, less coding

Ability to query complex aAbility to query complex applications and ability to pplications and ability to handle large and complex handle large and complex applications applications

Disadvantage Disadvantage Inability to handle compleInability to handle complex applications x applications

Low performance due to cLow performance due to complex query optimizatioomplex query optimization, inability to support largn, inability to support large-scale systems e-scale systems

Low performance in web aLow performance in web application pplication

Support from vendors Support from vendors It is considered to be highlIt is considered to be highly successful so the market y successful so the market size is very large but many size is very large but many vendors are moving towarvendors are moving towards ORDBMS ds ORDBMS

Presently lacking vendor sPresently lacking vendor support due to vast size of upport due to vast size of RDBMS market RDBMS market

All major RDBMS vendors All major RDBMS vendors are after this so has very gare after this so has very good future ood future

Slowly moveSlowly move

The other current ORDBMSs include OraThe other current ORDBMSs include Oracle8, from Oracle Corporation, and Univcle8, from Oracle Corporation, and Universal DB (UDB) from IBM. Also, Stonebraersal DB (UDB) from IBM. Also, Stonebraker point out that applications from Relaker point out that applications from Relational DBMSs (simple data with query) wtional DBMSs (simple data with query) will slowly move towards the Object-Relatill slowly move towards the Object-Relational DBMSs (complex data with query). ional DBMSs (complex data with query).

five architectural options given by Dfive architectural options given by Dr. Stonebraker r. Stonebraker

Supply plug-in code to make function calls to Supply plug-in code to make function calls to other applications. other applications.

Add separate API's and server subsystems to sAdd separate API's and server subsystems to support object functionality. upport object functionality.

Simulate specialized object-relational functioSimulate specialized object-relational functionality in a middleware layer. nality in a middleware layer.

Completely redesign the database engine. Completely redesign the database engine. Add a new object-oriented layer to support ricAdd a new object-oriented layer to support ric

h datatypes atop a proven relational database h datatypes atop a proven relational database engine. engine.

ConclusionConclusion

In spite of many advantages, ORDBMSs In spite of many advantages, ORDBMSs also have a drawback. The architecture also have a drawback. The architecture of object-relational model is not appropof object-relational model is not appropriate for high-speed web applications. Hriate for high-speed web applications. However, with advantages like large storaowever, with advantages like large storage capacity, access speed, and manipulage capacity, access speed, and manipulation power of object databases, ORDBMtion power of object databases, ORDBMSs are set to conquer the database markSs are set to conquer the database market. et.