Database Design Phase

  • Conceptual Design
    • Provide E-R Schema: description of requirements
    • Not an implementation model
  • Logical Design
    • Map to the data implementation model of DBMS
    • e.g., Relation Schema
  • Physical Design
    • Specify physical features of database
    • e.g., index

E-R Model

  • Entity: object
    • Entity Set: Set of entities of same type that shares the property.
  • Relationship: association among two or more entities
    • Relationship Set: Set of relationships
  • Attribute: properties of entity or relationship

E-R Diagram

Style 1

  • Rectangle: entity set
  • Diamond, connected with rectangles by line/arrow: Relationship set
    • Double diamond: identifying relationship set of weak entity set
    • Line: entity from opposite set can be mapped to many entities of this set
    • Arrow: entity from opposite set can be mapped to one entity of this set
  • Listed items in rectangle: attributes of entity
    • Underlined attributes: PK
    • Dash-underlined attributes: discriminator
    • Indentation: composite attribute
    • Curly bracket: multivalued attribute
    • (): derived attribute
  • Small box connected to diamond with dashed line: Attributes of relationship
  • Written text on line/arrow: role of the participating entity
  • Double line/arrow: total participation
  • Rectangle pointing other rectangle: generalization/specialization

Style 2

다른 점만 기술함

  • Attribute: pin
    • PK of entity set: black pin
    • composite attribute: pin on pin
  • Generalization/Specialization
    • Disjoint: colored arrow
    • Overlapping: uncolored arrow

Attribute Types

  • Simple vs Composite
    • Simple: atomic
    • Composite: composed of multiple parts
    • Why composite? It is more intuitive and better represents the requirement.
  • Single-Valued vs Multivalued
    • Single-valued: each entity has a single value for the attribute
    • Multivalued: each entity has multiple values for the attribute
  • Derived
    • Value of an attribute can be derived from other attributes or entities

Degree of Relationship Set

How many entity sets are involved is called degree of relationship set. Most relationships are binary, but not all (e.g., ternary).

A binary relationship whose involved entity sets are the same is sometimes called unary.

Mapping Constraints

Relationship Cardinality

A number of entities that can be mapped via a relationship. In other words, how many relationships each entity can participate.

  • one-to-one (1:1)
    • entity from A mapped to at most one entity from B (cardinality 1)
    • vice versa
  • one-to-many (1:m)
    • entity from A can be mapped to many entity from B (cardinality m)
    • entity from B mapped with at most one entity from A (cardinality 1)
  • many-to-one (m:1): opposite of one-to-many
  • many-to-many (m:n): no constraint

In diagram, arrow means one, line means many.

  • $A \leftarrow R - B$: one-to-many(1:m). c(A) = m, c(B) = 1.
  • $A \leftarrow R \rightarrow B$: one-to-one(1:1). c(A) = 1, c(B) = 1.
  • $A - R - B$: many-to-many(m:n). c(A) = n, c(B) = m.
  • $A - R \rightarrow B$: many-to-one(m:1). c(A) = 1, c(B) = m.

Participation

  • Total participation
    • every entity participates in at least one relationship
    • opposite: Partial participation

In diagram, double line means total participation, regardless of cardinality.

Alternative Notation

Express cardinality by upper & lower limits. Write on the line.

  • Max m (or *): many
  • Max 1: one
  • Min 0: partial participation
  • Min 1: total participation

Keys

  • Entity Set: Same as relational model
    • Superkey, Candidate key, Primary Key
    • Set of attributes whose values can distinguish entities from each other
  • Relationship Set: Combination of PKs from entity sets form superkey
    • Candidate key, Primary key: based on mapping constraints

Redundant Attributes

Relational Model에서와 다르게 entity간 relationship을 explicit하게 표현 가능. 따라서 foreign key에 해당하는 attribute는 redundant => should be removed

Weak Entity Set

An entity set that does not have sufficient attributes to form PK. PK에 redundant attribute가 포함되는 경우. 반대는 strong entity set.

Redundant attribute는 foreign key, 즉 다른 entity(identifying entity set)의 attribute이므로, 이와 연결해주는 identifying relationship set을 표시해야 함. (double diamond)

Weak entity set은 항상 total participation + strong entity set과 one-to-many.

PK - redundant attribute(s) = discriminator, or partial key

Specialization / Generalization

  • Specialization: entity set을 여러 그룹으로 쪼개어 sub entity set을 만드는 것
    • 상속: super entity set의 attribute들이 상속됨 + own attribute를 가질 수 있음
  • Generalization: 여러 entity set들을 합쳐 higher-level entity를 만드는 것
    • 디자인 과정에서 specialization의 반대로 진행되나, 결과만 놓고 보면 똑같음.

Types of Generalization

  • Disjoint: sub entity set들끼리 disjoint
  • Overlapping: 어떤 entity는 여러 sub entity set에 포함됨
  • Total: 모든 entity가 적어도 한 sub entity set에 포함됨
  • Partial: sub entity set에 포함되지 않는 entity가 있을 수 있음

(Logical Design) Reducing E-R Schema to Relation Tables

  • Entity Set => table
    • columns = attributes + one-to-many 당한 foreign key들
  • Relationship Set => table (only if many-to-many)
    • columns = attributes + participating entity들의 pk
    • one-to-one이나 one-to-many는 relationship을 foreign key로 표현
  • Generalization/Specialization => 2 options
    • foreign key로 표현(sub entity sets reference super entity set)
      • Partial and/or Overlapping generalization에 적합
    • sub entity set들 각각이 super entity set의 attribute를 가지도록 merge
      • Total and/or Disjoint generalization에 적합

Design Issues

Entity vs Attribute

  • Simple => attribute
  • Complex => entity
    • Entity는 다른 entity와 relationship을 가질 수 있음, 즉 독립적인 객체로서 생각할 수 있음
    • Multivalue attribute / shared value라면 entity로 만드는 것 고려 해볼만

Entity vs Relationship

  • Relationship: simple but limited
  • Entity: 독립적인 객체로 다른 entity와 relationship을 가질 수 있음

Binary vs n-ary

All n-ary relationships are representable by binary relationships, by adding an entity set that represents n-ary relationship set.

항상 모든 것을 binary relationship으로 표현하는 것은 diagram의 직관성을 오히려 해칠 수도 있다. 언제나 상황에 맞게 사용해야 함.

Generalization/Specialization

굳이 안해도 될 것인가 할 것인가

Design Quality

  • Completeness: all relevant features are represented
  • Correctness: concept of E-R model is properly used
    • Syntactic Correctness: concepts are properly defined
    • Semantic Correctness: Concepts are used according to their definition


Back

Categories:

Updated: