Transcript Query Processing, Resource Management and Approximate in …
4. Relational Databases
– –
•
Levels of Abstraction in data
defined by various “schema” levels
• Many
views
,
View 1 View 2 View 3
Conceptual (logical) schema
•
Conceptual Schema
Physical schema
.
Physical Schema
Views describe how users see data (possibly different data models for different views) Conceptual schema defines logical structure of entire data enterprise
–
Physical schema describes underlying files and indexes used.
ANSI schema model
Schemas are defined using Data Definition Languages or DDLs ;
data are modified/queried using Data Manipulation Languages or DMLs
.
Structure of a DBMS
• A typical DBMS has a layered architecture.
Query Optimization and Execution Relational Operators Files and Access Methods These layers must consider concurrency control and recovery • This is one of several possible architectures.
• Another with a little more detail on next slide.
Buffer Management Disk Space Management DB
Structure of a DBMS
QUERIES
from users (or Transactions or user-workload requests)
SQL (or some other User Interface Language)
QUERY OPTIMIZATION LAYER
Relational Operators (Select, Project, Join)
DATABASE OPERATOR LAYER
File processing operators (open,close file,read/write record
FILE MANAGER LAYER
(provide the file concept)
Buffer managment operators (read/flush page)
BUFFER MANAGER LAYER
Disk transfer operators (malloc, read/write block
DISK SPACE MANAGER LAYER
DB on DISK
DISK SPACE MANAGER
deals with space on disk offers an interface to higher layers (mainly the BUFFER MGR) consisting of: allocate/deallocate space; read/write block can be implement on a
raw disk system
directly, then it would likely access data as follows:
read block b of track t of cylinder c on disk d
or can use
OS file system
(OS file = sequence of bytes) then it would likely access data as follows:
read bytes b of file f
and then the Operating System file manager would translate that into
read block b of track t of cylinder c on disk d
most systems do not use the OS files system - for portability reasons, - to avoid OS file size peculiarities (limitations)
BUFFER MANAGER
partitions the main memory allocated to the DBMS into buffer page frames, brings pages to and from disk as requested by higher layers (mainly the FILE Mgr).
FILE MANAGER
supports the file concept to higher layers (DBMS file = collection of records and pages of records) supports access paths to the data in those files (e.g., Indexes).
Not all Higher level DBMS code recognizes/ uses page concept.
Almost all DBMS use the record concept, though.
DATABASE OPERATOR LAYER
implements physical data model operators (e.g., relational operators; select, project, join...)
QUERY OPTIMIZER
produces efficient execution plans for answering user queries (e.g., execution plans as trees of relational operators:
select, project, join, union, intersect
translated from, e.g., SQL queries). SQL is not adequate to answer all user-database questions, e.g., Knowledge workers working on Data Warehouses ask "what if" questions (On-Line Analytic Processing or OLAP) not retrieval questions (SQL)
CLUSTERING and Record Identification
Disk I/O minimization
is still main performance objective in designing a DBMS. Thus, clustering records on disk correctly is important.
CLUSTERING
= storing logically related records (those accessed together) physically close, to reduce disk access time.
If the workload typically requests sequential access by SUPPLIER#, then cluster: SUPPLIER-1-record close to SUPPLIER-2-record, ...
intra-file clustering
If the workload typically requests individual specific SUPPLIER-# followed by access to its shipment data, cluster: SUPPLIER-1-record , SUPLLIER-1's_shipment_data, SUPPLIER-2-record , SUPPLIER-2's_shipment_data, ...
inter-file clustering RID = Record Identifier
= permanently assigned record identifier (page#, record#)
RRN = Relative Record Number
= permanently assigned order number - usually an arrival order number
Overview of Database Design
Conceptual design
: What are the
entities
and
relationships
in the enterprise?
What information about these entities and relationships should be stored in the database?
What integrity constraints or business rules should be enforced? A database `schema’ Model diagram answers these question pictorially (Entity-Relationship or ER diagrams).
Then one maps the ER diagrams into a relational schema (using the Data Definition Language provided)
Entity Relationship Model
ssn name lot
Entity:
types. Real-world object type distinguishable from other object
Employee
An entity is described (in DB) using a set of
Attributes
.
– Each entity set has a
key
.(the chosen identifier attribute(s);
underlined
) – Each attribute has a
domain
.(allowable value universe)
ER Model (Cont.)
Relationship:
Association among two or more entities . E.g., Employee Jones works in Pharmacy department .
ssn name Employee lot since Relationships can have attributes too!
did dname budget ssn name lot Works_In Department
Degree=2 relationship between entities, Employees and Departments.
Employee super visor subor dinate Reports_To Degree=2 relationship between an entity and Itself? E.g., Employee Reports_To Employee.
Must specify the
“role”
of each entity to distinguish them.
Relationship Cardinality Constraints
(
many-to-many
) Works_In: An employee can work in many departments.
a dept can have many employees working in it.
• (
1-many
) e.g., Manages: • It may be required that each dept has at most 1 manager.
name Employee lot m name ssn Employee lot 1 since Works_In did n dname budget Department since Manages dname did budget m Department
• (
1-1
) Manages: In addition it may be required that each manager manages at most 1 department.
name ssn Employee lot 1 since Manages did dname budget 1 Department
Relationship Cardinality Constraints 1-to-1 1-to Many Many-to-1 Many-to-Many
Participation Constraints
Every department may have to have a manager?
This is an example of
total participation constraint
: the participation of Department in Manages is said to be total (vs. partial) .
ssn Employees lot Manages Works_In total Departments since
ISA (`is a’) Hierarchies
name ssn lot
We can use attribute inheritance to save repeating shared attributes.
hourly_wages
If we declare an ISA relationship among entity types, e.g., A ISA B (every instance of A entity is also an instance entity of entity B), then B entities “inherit” A entity attributes
hours_worked Employee Covering yes ISA Overlap allowed Hourly_Emp contractid Contract_Emp e.g., every Hourly_Emp ISA Employee every Contract_Emp ISA Employee Hourly_Emps and Contract_Emps can have their own separate attributes also.
Overlap constraints
: Can Joe be an Hourly_Emp and a Contract_Emp? (Allowed/disallowed)
Covering constraints (Yes/no)
: Does every Employee entity also have to be an Hourly_Emp or a Contract_Emp entity?
Relational Database: Working Definitions
•
Relational database:
a set of
relations
•
Relation:
made up of 2 parts:
–
Instance or occurrence
: a
table,
with rows and columns.
#Rows =
cardinality
, #fields =
degree
–
Schema or type
:
specifies name of relation & name, type of each attribute • Students(
sid
: string,
name
: string,
login
: string,
age
: integer,
gpa
: real).
• Strictly, a relation is a
set
of
tuples
but it is common to think of it as a table (sequence of rows made up of a sequence of attribute values)
Relational Query Languages
• A major strength of the relational model: supports simple, powerful
querying
of data. • Queries can be written intuitively (specifying what, not how), DBMS is responsible for evaluation
• The DBMS does your programming!
–
Allows a module called the optimizer to extensively re order operations (even combine similar operations from different concurrent requests), and still ensure that the answer does not change.
SQL Query Language
• One of the simplest languages on earth very English-like! Specify what, not how.
• E.g., SELECT attributes condition
What columns you want
FROM relations WHERE
What rows you want.
sid
• Find all 18 year old students ( a selection )
SELECT * FROM Students S WHERE S.gpa=3.4
sid name login 53666 Jones jones@cs age gpa 18 3.4
name login age gpa 53688 Smith smith@ee 18 3.2
53666 Jones jones@cs 18 3.4
•To find just names and logins ( a projection ), replace 1 st SELECT S.name, S.login
FROM Students S WHERE S.age=18 name login Jones jones@cs
line:
Querying Multiple Relations
(
Join,
implemented using nested loop – alternative 1 ) • What does the following query produce?
sid name SELECT FROM WHERE S.name, E.cid
Students S, Enrolled E login S.sid=E.sid AND 53666 Jones jones@cs age gpa 18 3.4
Where also used to combine (join) S & E E.grade=“A” sid cid 53831 Carnatic101 53831 Reggae203 grade C
suceeds But Select fails
B 53650 Smith smith@ee 18 3.2
53650 Topology112 A 53666 History105 B we get: S.name
Smith E.cid
Topology112
Destroying and Altering Relations (also DDL)
DROP TABLE Students
• Destroys the relation Students. The schema information
and
the tuples are deleted.
ALTER TABLE Students ADD COLUMN Year: integer
The schema of Students is altered by adding a new field; every tuple in the current instance is extended, e.g., with a
null
value in the new field.
Adding and Deleting Tuples
• Can insert a single tuple using:
INSERT INTO Students (sid, name, login, age, gpa) VALUES (53688, ‘Smith’, ‘smith@ee’, 18, 3.2)
Can delete all tuples satisfying some condition (e.g., name = Smith):
DELETE FROM Students S WHERE S.name = ‘Smith’
many powerful variants of these commands are available!
Views
• A view is a relation constructable from stored or base relations. Store a (actual tuples).
definition of it , rather than the instance CREATE VIEW Young ActiveStudents AS SELECT S.name, E.grade
FROM Students S, Enrolled E WHERE S.sid = E.sid
(name, grade) and S.age<21 Views can be dropped using the DROP VIEW command. How to handle DROP TABLE if there’s a view on the table?
DROP TABLE command has options to let user specify this.
• Views can be used to present necessary information (or a summary), while hiding details in underlying relation(s).
Integrity Constraints (ICs)
• IC: condition that must be true for any instance in the database; e.g., domain constraints.
• ICs are specified when (or after) relations are created.
– ICs are checked when relations are modified.
• A legal instance of a relation is one that satisfies all its ICs. – DBMS should not allow illegal instances.
– Avoids data entry errors, too!
Primary Key Constraints
• A set of fields is a key (strictly speaking, a candidate key) for a relation if it satisfies: 1. ( Uniqueness condition ) No two distinct tuples can have same values in the key (which may be a composite) 2. ( Minimality condition ) The Uniqueness condition is not true for any subset of a composite key.
– If Part 2 is false, it’s called a key ) superkey (for superset of a – There’s always at least one key for a relation, one of the keys is chosen (by DBA) to be the primary key , the primary record identification or lookup column(s) • E.g., sid is a key for Students. The set { sid, gpa } is a superkey.
Entity integrity
No column of the primary key can contain a null value.
Foreign Keys and Referential Integrity
• Foreign key to `refer’ to a tuple in another relation. (by listing the the primary key value in the second relation.) Like a `logical pointer’.
: A field (or set of fields) in one relation used • E.g. sid in ENROLL is a foreign key referring to Students (sid is the primary key of S) sid in – – If all foreign key constraints are enforced, a special integrity constraint, referential integrity , is achieved, i.e., no dangling references E.g., if Referential Integrity is enforced (and it almost always is) an Enrolled record cannot have a sid that is not present in Students ( students cannot enroll in courses until they register in the school )
Foreign Keys
• Only students listed in the Students relation should be allowed to enroll for courses.
Enrolled sid cid 53666 Carnatic101 53666 Reggae203 53650 Topology112 53666 History105 grade C B A B Students sid name login 53666 Jones jones@cs 53688 Smith smith@eecs 53650 Smith smith@math age gpa 18 18 19 3.4
3.2
3.8
Enforcing Referential Integrity
• Consider Students and Enrolled; sid foreign key that references Students.
in Enrolled is a • What should be done if an Enrolled tuple with a non existent student id is inserted? ( Reject it!
) • What should be done if a Students tuple is deleted?
– – – Also delete all Enrolled tuples that refer to it?
Disallow that deletion if an Enrolled tuple refers to it?
Set sid in Enrolled tuples that refer to it to a default sid ?
– (sometimes there is a “default default, e.g., set sid in Enrolled tuples to a special value null , denoting `not applicable’ if no other default is specified.)
Referential Integrity in SQL
• SQL supports all 4 options on deletes and updates.
– – Default action = NO ACTION (the violating
delete/update request is rejected
) CASCADE (also delete all tuples that refer to deleted tuple) CREATE TABLE Enrolled (sid CHAR(20), cid CHAR(20), grade CHAR(2), PRIMARY KEY (sid,cid), FOREIGN KEY (sid) REFERENCES Students ON DELETE CASCADE ON UPDATE SET NULL ) – SET NULL / SET DEFAULT (sets foreign key value of referencing tuple)
Where do ICs Come From?
• ICs are based on the semantics of the real-world enterprise that is being described in the database. I.e., the users decide semantics, not the DB experts!
Why?
• We can check a database instance to see if an IC is violated, but we can NEVER infer an IC by only looking at the data instances.
• An IC is a statement about all possible instances!
• An IC is a statement about all possible instances!
• It is not a statement that can be inferred from the set of currently existing instances. • If IC s were inferred from current instances, then when a relation is newly created and has, say, just 2 tuple, many, many ICs would be inferred (e.g., in
name login age gpa 53666 Jones jones@cs 18 3.4
53688 Smith smith@ee 18 3.5
the system might infer that students MUST be 18 or that names have to be 5 characters or worse yet, that gpa ranking must be the same as alphabetical name ordering!
Key and foreign key ICs are the most common.
• The next slides deals with the IC of choosing
Who decides primary key?
(and other design choices?) • The Database design expert?
– NO! Not in isolation, anyway.
– Someone from the enterprise who understands the data and the procedures should be consulted.
– The following story illustrates this point. CAST: – Mr. Goodwrench = MG (parts manager); – Pointy-headed Dbexpert = Ph D
Ph D I've looked at your data, and decided Part Number (P#) will be designated the primary key for the relation, PARTS(P#, COLOR, WT, TIME-OF-ARRIVAL).
MG You're the expert.
Ph D Well, according to what I’ve learned in school, P# should be the primary key, because IT IS the lookup attribute!
. . . later
MG Why is lookup so slow?
Ph D You do store parts in the stock room ordered by P#, right?
MG No. We store by weight!
When a shipment comes in, I take each part into the back room and throw it as far as I can. The lighter ones go further than the heavy ones so they get ordered by weight!
• Ph D But, but… weight doesn't have Uniqueness property!
the same weight end up together in a pile!
Parts with • MG No they don't. I tire quickly, so the first one I throw goes furthest.
• Ph D Then we’ll use a composite primary key, (weight, time-of arrival).
• MG We get our keys primarily from Curt’s Lock and Key.
• The point is: This conversation should have taken place during the 1
st
meeting.
An ER Example:
COMPANY is described to us as follows: 1. The company is organized into depts - each with a name, number, manager. - Each manager has a startdate. - Each department can have several locations. 2. Departments control projects - each with a name, number, location. 3. Each employee has a name, SSN, sex, address, salary, birthdate, dept, supervisor. - An employee may work on several projects (not necessarily all controlled by his dept) for which we keep hoursworked by project. 4. Each employee dependent has a name, sex, birthdate and relationship.
In ER diagrams, entities are represented in boxes: |EMPLOYEE| |DEPENDENT| |DEPT| |PROJECT| An attribute (or property) of an entity describes that entity. An ENTITY has a TYPE, including name and list of its attributes. ENTITY TYPE SCHEMA describes the common structure shared by all entities of that type. Project (Name, Num,Location, Dept) ENTITY INSTANCE = individual occurrence of an entity of a particular type at a particular time (Dome, 46, 19 Ave N & Univ, Athletics) (IACC, 52, Bolley & Centennial, C.S.) (Bean Res, 31, 12 Ave N & Federal, P.S.) . . . Entity Type does not change often - very static.
Entity instances get added, changed often - very dynamic
An ER Example continueed:
ATTRIBUTES are written next to Entity they describe, usually something like the following: Name-------------. Number-----------| Locations--------|--|DEPARTMENT| Manager----------| ManagerStartDate-' .--Name |--SSN |__EMPLOYEE__|----|--Sex |--Address |--Salary |--BirthDate |--Department |--Supervisor `--WorksOn Name-------------. Number-----------| Location---------|-|_PROJECT| ControlDepartment' .-Employee |-DependentName |_DEPENDENT|--|-Sex |-BirthDate `-Relationship
An ER Example:
CATEGORIES OF ATTRIBUTES: = COMPOSITE ATTRIBUTE = attributes that are subdivided into smaller parts with independent meaning.
e.g., Name attribute of Employee may be subdivided into FName, Minit, LName. Indicated: Name (FName, Minit, LName) Also, WorksOn may be a composite attr of Employee of Project and Hours: WorksOn (Project, Hours) SINGLE-VALUED ATTRIBUTE: one value per entry. MULTIVALUED ATTRIBUTE (repeating group) have multiple values per entry: eg, Locations (as an attribute of Department since a Department can have multiple locations) - Multivalued Attribute, use {Locations} - WorksOn may be a mutlivalued attr of Employee as well composite: {WorksOn (Project,Hours)} DERIVED ATTRIBUTE is an attribute whose value can be calculated from other attribute values. eg, Age calculated from BirthDate and CurrentDate. KEY ATTRIBUTE: Each value can occur at most once. (has the uniqueness property) Used to identify entity instances. We will * key attribute(s). ATTRIBUTE DOMAIN: Set of values that may be assigned (also called Value Set). Thus the Preliminary Design of Entity Types for COMPANY db is.
*Name-------------. *Number-----------| Locations--------|--|DEPARTMENT| Manager----------| ManagerStartDate-' .---Name |--*SSN |__EMPLOYEE__|----|---Sex |---Address |---Salary |---BirthDate |---Department |---Supervisor `---WorksOn *Name-------------. *Number-----------| Location---------|-|_PROJECT| ControlDepartment' .--Employee |-*DependentName |_DEPENDENT|--|--Sex |--BirthDate `--Relationship
An ER Example continued:
RELATIONSHIPS among entities express relationships among them: Relationships have RELATIONSHIP TYPEs (consisting of the names of the entities and the name of the relationship). A Relationship type diagram for a relationship between EMPLOYEE and DEPARTMENT called "WorksFor" is diagrammed: (in a roundish box) |EMPLOYEE|-( WorksFor )-|DEPARTMENT| RELATIONSHIP INSTANCEs for the above relationship might be, eg: ( John Q. Smith, Athletics ) ( Fred T. Brown, Comp. Sci.) ( Betty R. Hahn, Business ) . . . RELATIONSHIP DEGREE: Number of participating entities (usually 2) If an entity participates more than once in the same relationship, then ROLE NAMES are needed to distinguish multiple participations. eg, Supervisor, Supervisee in Supervision relationship - Called Reflexive Relationships. - Unnecessary if entity types are distinct. One decision that has to be made is to decide whether attribute or relationship is the appropriate way to model, e.g., "WorksOn". Above we modeled it as an attribute of EMPLOYEE {WorksOn(Project,Hours)} The fact that it is multivalued and composite (involving another entity, project) ssuggest that it would be better to model it as a relationship (i.e., it makes a very complex attribute!) WORKS_FOR(EMPLOYEE, DEPARTMENT)
An ER Example continued:
CONSTRAINTS ON A RELATIONSHIP CARDINALITY CONSTRAINT can be 1-to-1 many-to-1 1-to-many or many-to-many 1 to 1: MANAGES(EMPLOYEE, DEPARTMENT) Many to 1: WORKS_FOR(EMPLOYEE, DEPARTMENT) Many to Many: WORKS_ON(EMPLOYEE, PROJECT) Each manager MANAGES 1 dept Each dept is MANAGED-BY 1 manager Each employee WORKS_FOR 1 dept Each dept is WORKED_FOR by many emps Each employee WORKS_ON many projects Each project is WORKED_ON by many employees PARTICIPATION CONSTRAINT (for an entity in a relationship) can be Total, Partial or Min-Max Total: Every EMPLOYEE WORKS_FOR some DEPARTMENT Partial: Not every EMPLOYEE MANAGES some DEPT RELATIONSHIP can have ATTRIBUTES (properties) as well: eg, Hours for WORKS_ON Relationship, Manager_Start_Date in MANAGES relationship.
An ER Example continued:
6 RELATIONSHIPS; CARDINALITY ---------- 1:1 1:many many:many 1:many RELATIONSHIP ----------- MANAGES WORKS_FOR WORKS_ON CONTROLS (role names, if any, above) ATTRIBUTES ------ (participation below) (EMPLOYEE, DEPARTMENT) partial total (DEPARTMENT, EMPLOYEE) total total (EMPLOYEE, PROJECT) total total (DEPARTMENT, PROJECT) partial total Reflexive relationship with role names --------------.---------. supervisor supervisee 1:many SUPERVISION (EMPLOYEE, EMPLOYEE) partial partial 1:many DEPENDENTS_OF ( EMPLOYEE, DEPENDENT) partial total
An ER Example continued:
COMPANY Entity-Relationship Diagram (showing the Schema) (double connecting lines means "total" while single line means partial participation.) ( MANAGES ) 1|| |1 || (WORKS_FOR) | *Name-----------. || 1|| many|| | *Number---------| ( CONTROLS ) || || || / {Locations}-----|- DEPARTMENT // / number_employees' /1 // / .----' // / // / many // / || // / || (SUPERVISE) // / || | | // / || 1| |many // / || 'er| |'ee // / || |____|_______//_____/ .Name(FN,Mi,LN) || |_EMPLOYEE__________|---|-*SSN || // | |-Sex || // | |-Address || Hours-. // | |-Salary || | /many | `-BirthDate \\ (WORKS_ON) | \\ \\ 1| many| | \\ || ( Dependent_0f ) \\ || |many *Nane-. \\_______||___ || *Numb-|--| PROJECT | || Locatn' || || *DependentName---. .
Sex--------------|--|| DEPENDENT || BirthDate--------| Relationship-----'