Introduction

CPS 116 Introduction to Database Systems

Course goals

Random things you might do (for fun or profit) after taking this course
Develop your own database-driven Web sites (like Amazon, eBay, etc.) Be a power user of commercial database systems Upgrade your Web sites with XML Explain to friends why MySQL is not a real database system without InnoDB or Berkeley DB support

Course roadmap
Relational databases
Relational algebra, database design, SQL, application programming

XML
Data model and query languages, application programming, interplay between XML and relational databases

Database internals
Storage, indexing, query processing and optimization, concurrency control and recovery

Research topics
Data warehousing and mining, stream data processing, etc.

What is a database system?

From Oxford Dictionary: Database: an organized body of related information Database system, DataBase Management System (DBMS): a software system that facilitates the creation and maintenance and use of an electronic database

What do you want from a DBMS?

Keep data around (persistent) Answer queries (questions) about data Update data Example: a traditional banking application
Each account belongs to a branch, has a number, an owner, a balance, Each branch has a location, a manager, Persistency: Homer will be pretty upset if his balance disappears after a power outage Query: Whats the balance in Homer Simpsons account? Modification: Homer withdraws $100

Sounds simple!
1001#Springfield#Mr. Morgan ... ... 00987-00654#Ned Flanders#2500.00 00123-00456#Homer Simpson#400.00 00142-00857#Montgomery Burns#1000000000.00 ... ...

ASCII file Accounts/branches separated by newlines Fields separated by #s

Query
1001#Springfield#Mr. Morgan ... ... 00987-00654#Ned Flanders#2500.00 00123-00456#Homer Simpson#400.00 00142-00857#Montgomery Burns#1000000000.00 ... ...

Whats the balance in Homer Simpsons account? A simple script

Scan through the accounts file Look for the line containing Homer Simpson Print out the balance

Query processing tricks

Tens of thousands of accounts are not Homers

And the list goes on

What happens when the query changes to: Whats the balance in accounts 00142-00857?

Observations

Tons of tricks (not only in query processing, but also in storage, concurrency control, recovery, etc.) Different tricks may work better in different usage scenarios (example?) Same tricks get used over and over again in different applications We need a library, or better yet, a server (to support sharing, backup, etc.)

The birth of DBMS 1

10

(Pretty drawing stolen from Hans-J. Scheks VLDB 2000 slides)

The birth of DBMS 2

11

(Pretty drawing stolen from Hans-J. Scheks VLDB 2000 slides)

The birth of DBMS 3

12

(Pretty drawing stolen from Hans-J. Scheks VLDB 2000 slides)

Early efforts
Factoring out data management functionalities from applications and standardizing these functionalities is an important first step
CODASYL standard (circa 1960s) Bachman got a Turing award for this in 1973

13

But getting the abstraction right (the API between applications and the DBMS) is still tricky

CODASYL
Query: Who have accounts with 0 balance managed by a branch in Springfield? Pseudo-code of a CODASYL application:
Use index on account(balance) to get accounts with 0 balance; For each account record: Get the branch id of this account; Use index on branch(id) to get the branch record; If the branch records location field reads Springfield: Output the owner field of the account record.

14

Programmer controls navigation: accounts branches

How about branches accounts?

Whats wrong?
When data/workload characteristics change
The best navigation strategy changes The best way of organizing the data changes

15

With the CODASYL approach

To write correct code, application programmers need to know how data is organized physically (e.g., which indexes exist) To write efficient code, application programmers also need to worry about data/workload characteristics Cant cope with change!

The relational revolution (1970s)

A simple data model: data is stored in relations (tables) A declarative query language: SQL
SELECT Account.owner FROM Account, Branch WHERE Account.balance = 0 AND Branch.location = Springfield AND Account.branch_id = Branch.branch_id;

16

Programmer specifies what answers a query should return, but not how the query is executed DBMS picks the best execution strategy based on availability of indexes, data/workload characteristics, etc. Provides physical data independence

Physical data independence

17

Applications should not need to worry about how data is physically structured and stored Applications should work with a logical data model and declarative query language Leave the implementation details and optimization to DBMS The single most important reason behind the success of DBMS today
And a Turing Award for E. F. Codd

Modern DBMS features

Persistent storage of data Logical data model; declarative queries and updates physical data independence
Relational model is the dominating technology today XML is a hot wanna-be

18

What else?

DBMS is multi-user
Example
get account balance from database; if balance > amount of withdrawal then balance = balance - amount of withdrawal; dispense cash; store new balance into database;

19

Homer at ATM1 withdraws $100 Marge at ATM2 withdraws $50 Initial balance = $400, final balance = ?

Final balance = $300

Homer withdraws $100:
read balance; $400

20

Marge withdraws $50:

read balance; $400 if balance > amount then balance = balance - amount; $350 write balance; $350

if balance > amount then balance = balance - amount; $300 write balance; $300

Final balance = $
Homer withdraws $100:
read balance; read balance; if balance > amount then balance = balance - amount; write balance; if balance > amount then balance = balance - amount; write balance;

21

Marge withdraws $50:

Concurrency control in DBMS

Appears similar to concurrent programming problems?
But data not main-memory variables

22

Appears similar to file system concurrent access?

Approach taken by MySQL in the old days
(fun reading: http://openacs.org/philosophy/why-not-mysql.html)

But want to control at much finer granularity

Or else one withdrawal would lock up all accounts!

Recovery in DBMS
Example: balance transfer
decrement the balance of account X by $100; increment the balance of account Y by $100;

23

Scenario 1: Power goes out after the first instruction Scenario 2: DBMS buffers and updates data in memory (for efficiency); before they are written back to disk, power goes out Log updates; undo/redo during recovery

Summary of modern DBMS features

Persistent storage of data Logical data model; declarative queries and updates physical data independence Multi-user concurrent access Safety from system failures Performance, performance, performance

24

Massive amounts of data (terabytes ~ petabytes) High throughput (thousands ~ millions transactions per minute) High availability ( 99.999% uptime)

Major DBMS today

Oracle IBM DB2 (from System R, System R*, Starburst) Microsoft SQL Server NCR Teradata Sybase Informix (acquired by IBM) PostgreSQL (from UC Berkeleys Ingres, Postgres) Tandem NonStop (acquired by Compaq, now HP) MySQL and Microsoft Access

25

Modern DBMS architecture

Applications

26

DBMS OS Disk(s)

OS layer is bypassed by performance and safety Many details will be filled in the DBMS box

People working with databases

27

End users: query/update databases through application user interfaces (e.g., Amazon.com, 1-800-DISCOVER, etc.) Database designers: design database schema to model aspects of the real world Database application developers: build applications that interface with databases Database administrators (a.k.a. DBAs): load, back up, and restore data, fine-tune databases for performance DBMS implementors: develop the DBMS or specialized data management software, implement new techniques for query processing and optimization

Course information
Book
Database Systems: The Complete Book, by H. Garcia-Molina, J. D. Ullman, and J. Widom Get the value-pack edition with free access to Gradiance when it comes out (check Web site for updates)

28

Web site
http://www.cs.duke.edu/courses/fall04/cps116/

Course information; tentative syllabus and reference sections in GMUW; lecture slides, assignments, programming notes

Blackboard: for grades only Newsgroup: duke.cs.cps116

Preferred news server: news.cs.duke.edu (news.duke.edu also works but with a lag)

Course load
Four homework assignments (35%)
Include written and programming problems as well as online exercises on Gradiance

29

Course project (25%)

Details to be given in the third week of class

Optional presentation (8.75%; replaces lowest homework grade)

Sign-up sheet to be given in the third week of class

Midterm and final (20% each)

Open book, open notes Final is comprehensive, but emphasizes the second half of the course

10