ObjectOriented Databases

1
Object-Oriented Databases

• Andy Stafford
• Kelly Traw
• Max Podlesny
• MyoungSoo Park

2
Overview

• Introduction
• Problems with RDBMS
• Motivation for OO Data Model
• Review of OO concepts
• OODB design
• Data Storage and Access
• Object Models and Specification Languages
• Object-Relational Databases
• Strengths of OODBs
• Weaknesses of OODBs

3
RDBMS Weaknesses

• Fragmentation of real word entities through
  normalization
• Lacks good support for integrity constraints
• Fixed set of operations
• Lacks seamless integration between DBMS and host
  language, i.e. impedance mismatch

4
OODB Motivation

• Relation Model not powerful enough for many
  advanced database applications
• Computer Aided Drafting (CAD)
• Many types, dynamic design, parallel development
  (cooperative engineering)
• Office Information Systems (OIS) and multimedia
  systems
• Complex data storage (e.g. photographs,
  spreadsheets, voice commentary, hand drawn
  sketches, etc ..)

5
The OO debate

• Agreement in database field that RDBMSs are
  inadequate for complex applications
• RDBMS proponents argue support for such
  applications can be handled by extensions to
  relational model
• OODB proponents argue an entirely new paradigm is
  necessary to support these complex applications

6
Object Oriented Database definition

• A persistent and sharable collection of objects
  defined by an OODM
• Another definition, abstract data types
  inheritance object identity database
  capabilities
• No Universal agreement on OODBM!

7
OODBMS Manifesto

• Lays out thirteen required features of an OODBMS
  including
• Standard OO characteristics
• E.g. encapsulation, types, classes, polymorphism,
  etc..
• DBMS characteristics
• Persistence
• Scalable
• Concurrency
• Fault tolerance
• Simple query support

8
The Object in the OODM

• Similar to notion of Object in OOP object has
  state and behavior
• Defined by classes
• Constructors / Destructors
• Operator overloading
• Inheritance / Multiple inheritance
• Polymorphism and Dynamic (late) Binding
• Additionally, each object has unique Object
  Identifier (OID)
• System generated
• No dangling references
• Enforces referential integrity
• PK concept in RDBMS not sufficient
• Fast and efficient

9
OO Database Design

• OO Data Modeling (OODM) similar to EER (CDM)
  model
• Relationships represented by reference attributes

10
UML and OO Database Design

• UML can be used to aid in OO Database Design
• Use Case Diagrams
• Class Diagrams
• Sequence Diagrams
• State Diagrams

11
Persistence in OODBMS

• Checkpointing
• Saves entire address space to disk to secondary
  storage
• May contain data that doesnt need to be
  persisted
• Serialization
• Write objects to flat-file
• Not incremental writing large object expensive
• Explicit paging
• Reachability-based
• Allocation-based

12
Data Storage and Access

• RDBMSs two level storage model
• Application storage model in main virtual
  memory
• Database storage model on disk

Transformation
Main Memory
Secondary Storage
13
Data Storage and Access

• OODBMS uses single level storage concept

Main Memory
Secondary Storage
14
Data Storage and Access

• Resident Object Table (ROT)
• Easy implementation, potentially inefficient
• Pointer Swizzling
• More complex, more efficient

15
Object Management Group (OMG)

• Created the Object Management Architecture (OMA)
• Object Model (OM)
• Object Request Broker (ORB)
• Object Services
• Common Facilities
• Common Object Request Broker Architecture (CORBA)
• Interface Definition Language (IDL)
• Model-Driven Architecture (MDA)

16
Object Data Standard ODMG 3.0

• Proposed by the Object Data Management Group
  (ODMG)
• (Sun Microsystems, eXcelon Corp., Objectivity
  Inc., POET Software, Computer Associates, Versant
  Corporation)
• Four main parts
• Object Model
• Object Definition Language
• Object Query Language
• C, Java, and Smalltalk language bindings

17
Object Data Standard

• Object Definition Language
• Extends Interface Definition Language
• Object Query Language
• Expressions
• Many types - elementary, construction, atomic,
  object, collection, conversion, and indexed
  collection expressions
• Queries
• ODMG Language Bindings
• Object Manipulation Language (OML)

18
Applied Object Databases

• ObjectStore a commercial OODBMS
• Offers real-time data management, distributed
  data caching, and complex event processing
• Delta Airlines, Northrop-Grumman, Amazon.com,
  Federal Express
• http//www.objectstore.com/index.ssp
• Caché - a post-relational database
• Supports both SQL and object data access
• Caché object model is based upon the ODMG
  standard
• Fast transaction processing, scalability, and
  capability to execute real-time queries against
  transactional data
• Available for Windows, OpenVMS, Mac OS X, Linux
  and UNIX platforms
• http//www.intersystems.com/cache/index.html

19
Object-Relational Database Systems

• RDBMS and OODBMS hybrid
• Advantages
• Reuse
• Sharing
• Disadvantages
• Complexity
• Costs

20
RDBMS

• Decomposition of objects to primitive level
• Data objects stay in a rudimentary state
• - Number, Character
• Pointers do not exist ( common in Hierarchical
  and Network DB)
• Data Access is enabled by Primary and Foreign Key

21
RDBMS

• The major advantage of declarative structure
• The actual navigation path is hidden from users
• SQL optimizer determines the path
• Not supported
• When working with collection types and complex
  types, key can not reference the objects

22
Transition to ORDBMS

• Consider RDBMS first
• Add the feature of object-oriented processing to
  RDBMS
• Redesign RDBMS structure to support multimedia
  function
• Storage techniques for handling objects

23
To Enable Object Functionality

• Two Important points
• Logical design aspects of schema
• - Data type
• - How to access data
• The physical implementation of the logical schema
• - SQL semantic support object

24
Object Functionality in ORDBMS

• Behavior
• - A method a function of a particular class
• Collection type
• An aggregate object
• - a data-type definition composed of many
  subtypes coupled with behaviors
• A nested table a derived table

25
Object Functionality in ORDBMS

• Encapsulation
• Defining Class
• - Data ? attribute
• - Function ? method
• Polymorphism
• - Each class can have deferent objects for
  deferent responses to same request
• - Function overloading in C
• Inheritance
• - Subclassing
• - An object inherits a certain set of
  attributes from another object

26
ORDBMS Features

• Constituent element for object orientation
• - User-definition types
• - Type constructor for row types, reference
  type and collection types
• - User-definition function and procedures
• - Support for large objects ( Binary Large
  objects and Character Large Objects )

27
Benefit of reference type

• Pointer to repeating values
• - Simplify DB design
• Accessing non-database objects in a flat file
• - Used as a means of effective multimedia
  data accessing
• Data relationship without foreign key
• - Instead of slow SQL JOIN statements

28
Traditional System Analysis

• Data Flow Diagram (DFD)
• - A description of all processes, the data
  flow between them and data stores.
• Data Dictionary
• - A description of the logical data items
• Process logic specification
• - A description of all functional primitive
  processes

29
Object-oriented analysis

• Traditional structured analysis AND
• A list of all objects
• A list of all services (behaviors) of processes
• A list of associating the primary services with
  each object
• A description of relationships in the system
• A behavior script for each object
• A classification for each object and the object
  relationship

30
Choosing a Data Model

• When You have to decide to choose a DBMS package
  related to Object- oriented paradigm
• The third options
• Neither ORDMS nor OODBMS
• Used for avoiding complexity and high cost

31
Component Object Model (COM)

• Microsoft's framework for developing and
  supporting program component objects
• Equivalent capabilities to features in CORBA
• A framework for the interoperation of distributed
  objects in a network

32
The Third Option (Example)
Request
Response
ATL COM Server
ODBC
SQL Server
33
Why choose OODB?
34
OODBs allow users to define abstractions

• Ability to define new abstractions
• Ability to control the implementation of
  abstractions
• Allowance the definition of objects as aggregates
  of other objects
• Allowance to have aggregations be nested at
  several levels

35
OODBs allow users to define abstractions

• Ability to make properties be non-primitive
  objects as values, allowing deeply nested object
  structures
• Use of multi valued properties to express complex
  data structures.
• It is not the case in the RDB model where one
  need to use additional relations and joins

36
OODBs facilitate development of some relationships

• The feature of inverse relationships to express a
  mutual reference between two objects (a binary
  relationship)
• This feature ensures referential integrity by
  establishing corresponding reference as soon as a
  reference is created
• Possibility to automatically propagate deletion
  via these references

37
OODBs eliminate need for user defined keys

• OID is generated automatically by the system and
  that guarantees uniqueness to each object
• Elimination the need for user to define keys in
  the OODB model

38
OODBs eliminate need for user defined keys

• Elimination the need for user to define keys in
  the OODB model
• the OID cannot be modified by the application
• the notion of object identity provides a separate
  and consistent notion of identity, independent of
  how an object is accessed or modeled with
  descriptive data
• two objects are different if they have different
  OID even if they have the same structures and the
  same values for all properties
• user-defined keys are used to provide uniqueness
  of objects in RDBs

39
Development of equality predicates

• In the RDB model equality is always based only on
  values of attributes
• OODBs have different types of equality

40
Development of equality predicates

• OODBs have different types of equality
• Identity equality of objects. Two objects, S1 and
  S2 are equal if they have the same OID.
• Value equality of objects. This can be determined
  in two ways
• Two primitive objects are equal if they have the
  same value.
• Two non-primitive objects are equal if they have
  the same number of properties, and if, for any
  property Pi of S1 there exists a property Pi of
  S2 that is equal in value.
• Value equality of properties.
• Identity equality of properties.

41
OODBs reduce need for Joins

• The capability of navigating through object
  structures and the resulting path expressions in
  object attributes provides two types of join
• the implicit join, derived from the hierarchical
  nesting of objects
• the explicit join, which is similar to the
  relational join where two objects are explicitly
  compared by using either the value or the
  identity equality

42
Performance gain using OODBs

• The value of an attribute of an object X, whose
  domain is another object Y, is the object
  identifier (OID) of the object Y
• If an application has already retrieved object X
  and now would like to retrieve object Y, the
  database system could retrieve object Y by
  looking up its OID.
• RDBs do not support OID so this cannot be made so
  easily

43
Performance gain using OODBs

• Most OODBs convert the OIDs stored in an object
  to memory pointers when the object is loaded into
  memory
• RDBs cannot store memory pointers to other tuples
  because of absence of OIDs
• the facility to navigate through memory-resident
  objects is a fundamentally absent feature in RDBs
• the performance drawback that results from it
  cannot be neutralized by simply having a large
  buffer space in memory

44
Support for versioning or long-duration
transactions

• Few OODBs offer versioning and long-duration
  transactions, though with limited facilities
• Versioning and long-duration transactions are
  missing in RDBs

45
Development of Object Algebra

• Object algebra defines five fundamental
  object-preserving operators
• Union. Returns objects that are in sets P or Q or
  both.
• Difference. Returns the set of objects that are
  in set P and not in set Q.
• Select. Returns a subset of an input set.
• Generate. Generates objects from those in the
  input sets.
• Map. Returns a set of objects resulting from each
  sequence application.
• Other operators like intersection may be defined
  from these fundamental operators.

46
What OODB weaknesses exist?
47
Interoperability between RDBs and OODBs

• Absence of sufficient compatibility with RDBs
• Need of a migration path to allow the coexistence
  and the gradual migration from the current
  products to new products
• Absence of unified architectures of RDBs and
  OODBs
• Absence of unified data models of RDBs and OODBs

48
Minimal query optimization

• One of the biggest problems
• Complexity of the optimization of OODBs queries
  because of the presence of the additional
  complexity of the object-oriented data model
• additional data types
• changing variety of types
• complex objects, methods and encapsulation
• the use of nested structures
• object identity
• Todays OODBs offer rather simple optimization
  strategies.

49
Lack of standard query algebra

• Other big problem
• Several different formal query languages of
  algebras and calculi have been proposed
• Absence of analogous, intuitive correspondence
  between object algebra operators and physical
  system primitives in OODBs
• Close correspondence between algebra operations
  and low level primitives of the physical system
  in RDBs

50
Lack of query facilities

• Only few systems provide significant query
  facilities, but the query language is not ANSI
  SQL compatible.
• No object query standard

51
Security

• Most OODBs do not support authorization
• RDBs have a good support mechanism

52
Little support for complex objects

• Absence of full functionality of complex objects
• no predefined generic operations exploiting
  different reference semantics

53
No support for dynamic class definition changes
with OODBs

• Most OODBs do not allow dynamic changes to the
  database schema, such as
• adding a new attribute or method to a class
• adding a new superclass to a class
• dropping a superclass from a class
• adding a new class
• dropping a class
• RDBs allow the user to change the database schema
  dynamically using the ALTER command

54
Limited support for consistency constraints in
OODBs

• Absence of mechanisms to declare
• key properties of attributes (for example, an
  attribute of a class cannot be declared the
  primary key of the class) or uniqueness
  constraints
• explicit consistency constraints
• pre and postconditions of methods
• Explicit consistency constraints would be more
  easily accessible for inspection and modification

55
Limited performance tuning capability in OODBs

• Most of the OODBs offer limited capabilities for
  parameterized performance tuning

56
Limited performance gain over RDBs

• No performance advantages over RDBs while meeting
  the following requirements
• bulk database loading
• creation, update, and delete of individual
  objects
• retrieval from a class of one or more objects
  satisfying certain search conditions
• joins of more than one class
• transaction commit

57
Other features that OODBs do not support

• Triggers
• Meta data management features
• Constraints such as UNIQUE and NULL
• No support for views in OODBs
• OODBs do not support views
• The development of an object-oriented view
  capability is complicated by object identity.
• There is also the argument that data
  encapsulation and inheritance make explicit view
  definitions unnecessary

58
References

• Sikha Bagui, Achievements and weaknesses of
  Object-Oriented Databases, Journal of Object
  Technology, Vol. 2, No. 4, July-August 2003
• Piotr Habela and Kazimierz Subieta, Overcoming
  the Complexity of Object-Oriented DBMS Metadata
  Management