Database Application Design

(C) 2000, The University of Michigan

1

Database Application Design

Handout #8

February 25, 2000

(C) 2000, The University of Michigan

2

Course information

• Instructor: Dragomir R. Radev ([email protected])

• Office: 305A, West Hall

• Phone: (734) 615-5225

• Office hours: Thursdays 3-4 and Fridays 1-2

• Course page: http://www.si.umich.edu/~radev/654w00

• Class meets on Fridays, 2:30 - 5:30 PM, 311 WH

(C) 2000, The University of Michigan

3

Managing multi-user databases(cont’d)

(C) 2000, The University of Michigan

4

Concurrency control

• Lax and strict policies

• Atomic transactions (LUWs = logical units of work)– Example: customer+salesperson

• Concurrent transaction processing: interlocking

• Lost update problem

(C) 2000, The University of Michigan

5

Example

• User A:– Read item 100– Reduce by 5– Write item 100

• User B:– Read item 200– Reduce by 3– Write item 200

(C) 2000, The University of Michigan

6

Resource locking

• Locks: implicit, explicit

• Example: two users

(C) 2000, The University of Michigan

7

Example

• User A:– Lock item 100– Read item 100– Reduce by 5– Write item 100

• User B:– Lock item 100– Read item 100– Reduce by 3– Write item 100

(C) 2000, The University of Michigan

8

Example (cont’d)1. Lock item 100 for A

2. Read item 100 for A

3. Lock item 100 for B; cannot

4. Decrease 100 by 5

5. Write item 100 for A

6. Release A’s lock on 100

7. Lock item 100 for B

8. Read item 100 for B

9. Decrease item 100 by 3

10. Write 100 for B

11. Release B’s lock on 100

(C) 2000, The University of Michigan

9

Resource locking

• Serizalizable transaction– 2PL: growing phase, followed by a shrinking

phase

• COMMIT and ROLLBACK

• DEADLOCKS

(C) 2000, The University of Michigan

10

Transaction isolation levels

• Exclusive use

• Repeatable read: mix of shared and exclusive locks

• Dirty read: for reports which don’t necessarily need to contain the latest data

(C) 2000, The University of Michigan

11

Cursor types

• Forward only: changes made to earlier records are hidden

• Static: any changes are hidden

• Dynamic: all changes are visible

(C) 2000, The University of Michigan

12

Database recovery

• Reprocessing: uses database saves

• Rollback/Rollforward : uses transaction logs, before-images, and after-images

(C) 2000, The University of Michigan

13

Database security

• Users, groups, permissions, objects

• Permissions:– CONNECT: ALTER SESSION, CREATE

TABLE, CREATE VIEW

(C) 2000, The University of Michigan

14

Application security

• Usually done on the Web server

• ASP script modifies SQL statement:

SELECT *FROM EMPLOYEE<% WHERE EMPLOYEE.Name “=SESSION(“EmployeeName”)”%>

(C) 2000, The University of Michigan

15

Sharing enterprise data

(C) 2000, The University of Michigan

16

Enterprise DB architectures

• Teleprocessing systems

• Client-server systems

• File-sharing systems

• Distributed database systems: vertical and horizontal fragmentation

(C) 2000, The University of Michigan

17

Comparing distributed DB architectures

Unified database

Distributed databases

Single Nonpartitioned Nonreplicated

Partitioned Nonreplicated

Nonpartitioned Replicated

Partitioned Replicated

Increased security risk

Increased difficulty of control

Increased cost/complexity

Increased availability

Increased flexibility

Increased independence

Increased parallelism+

+

+

+

+

+

+

-

-

-

-

+

+

+

(C) 2000, The University of Michigan

18

Problems in downloaded databases

• Coordination

• Consistency

• Access control

• Computer crime

(C) 2000, The University of Michigan

19

On Line Analytic Processing (OLAP)

• Hypercubes, axes, dimensions, slices

• Values of a dimension are called members

• Levels: hierarchical organization: e.g., date, month, year

• CROSSJOIN ({Existing Structure, New Construction}, {California.Children, Nevada})

20

OLAP SQLCREATE CUBE HousingSalesCube ( DIMENSION Time TYPE TIME, LEVEL Year TYPE YEAR, LEVEL Quarter TYPE QUARTER, LEVEL Month TYPE MONTH, DIMENSION Location, LEVEL USA TYPE ALL, LEVEL State, LEVEL City, DIMENSION HousingCategory, DIMENSION HousingType, MEASURE SalesPrice, FUNCTION AVG, MEASURE AskingPrice, FUNCTION AVG)

(C) 2000, The University of Michigan

21

KDD: Data Mining

(C) 2000, The University of Michigan

22

Association rules

• X Y

• 65% of all customers who buy beer and tomato sauce also buy pasta and chicken wings

• Support (X)

• Confidence (X Y) = Support(X+Y) / Support (X)

(C) 2000, The University of Michigan

23

Object-oriented data processing

(C) 2000, The University of Michigan

24

Introduction

• OOP objects: encapsulated structures with attributes and methods

• Interface + implementation

• Inheritance

• Polymorphism

• Transient and persistent objects

(C) 2000, The University of Michigan

25

Final project guidelines

26

ChecklistIntroductionUser interviews/needs: table, reports, queries, formsInitial data modelER modelDecompositionSQL codeDocumentationEvaluation, Future workScheduleSustainabilitySnapshotsPresentationDemo

(C) 2000, The University of Michigan

27

Grading

• Project: 40%- design 10%- implementation 10%- documentation 10%- presentation+demo 10%

(C) 2000, The University of Michigan

28

Readings for next time

• Kroenke– Chapter 14: Sharing Enterprise Data

– Chapter 17: Object-Oriented Database Processing

• YRK (optional)– Chapter 14: Java and JDBC