Foundations of Databases

Transcription

1 Foundations of Databases (Datenbanken I) Prof. Dr. Torsten Grust U Tübingen, Database Systems torsten.grust@uni-tuebingen.de Summer Term 2011 Welcome to the completely rectangular world of Relational Database Management Systems (RDBMSs). What this course is about: Convince you that there is more to database technology than just open-file(), read()/write(), close-file(). Make you see how versatile the strictly tabular data model supported by relational databases can be. Make you best friends with SQL, the principal language spoken by relational database systems. We will encounter a healthy mix of good, clean theory and highly relevant CS practice knowledge of RDBMSs and SQL makes you sexy (in a sense). 2

2 Administrativa (1) 3 Lectures Time slot Monday, Tuesday, Room Sand 6/7, gr. Hörsaal Sand 6/7, gr. Hörsaal Practice Time slot Thursday, Room Sand 6/7, gr. Hörsaal Will this be a good fit for most of you? Please speak up. Administrativa (2) 4 End-term Exam 90 mins. examination on Monday, July 18th, You may bring a A4 double-sided hand-written cheat sheet. Passing earns you 6 ECTS. Assignments and Grading We will distribute, collect, and grade weekly assignments. You may and you should work in teams of two. Scoring 2 /3 of the overall points in the assignments earns you an additional 2 ECTS.

3 Administrativa (3) 5 Course web home teaching/ss11/db1 Download slides (PDF please bring a print-out and take notes) Download assignments, sample tabular data, SQL queries, other code snippets,... Please check now and then (... assignment unsolvable as given...,... no lecture on..., etc.) Contact information Just drop by in our offices, send first if you require specific help/longer attention. These Slides 6 Examples Definitions Code snippets Quizzes A specific slide set suitable for printing (lighter colors,... ) will be available on the course web home.

4 7 Read a Book, Write some SQL Text book (any introductory book is probably fine ask me) Alfons Kemper, Andre Eickler Datenbanksysteme Eine Einfu hrung Oldenbourg Verlag 6th or 7th ed. Install IBM DB2 V9.7 Express-C : Full-featured, fast, freely available. We will bring it with us for almost any lecture.

5 ActiveRecord (Ruby on Rails) 9 If time permits, we will close the course with an introduction to and overview of ActiveRecord. ActiveRecord enables a truly seamless embedding of database access and query functionality into programming and scripting languages (here: Ruby). You write Ruby fragments, the ActiveRecord framework generates equivalent SQL commands for you. ActiveRecord is the glue between relational databases and front-end web applications, usually developed using Ruby on Rails. Introduction 10 After completing this chapter, you should be able to: explain basic notions: database state, schema, query, update, data model, DDL, DML, explain the role of the DBMS, explain data independence, declarativity, and the three schema architecture, name different classes of users of a database application system, name some DBMS tools.

6 Introduction 11 Overview 1. Basic Database Notions 2. Database Management Systems (DBMS) 3. Programmer s View, Data Independence 4. Database Users and Database Tools Task of a Database (1) 12 What is a database? Difficult question. There is no precise and generally accepted definition. Naïve approach: The main task of the database system (DBS) is to answer certain questions about a subset of the real world, e.g. Questioning a DBS Which homeworks has Ann Smith completed? Database System 1 2

7 Task of a Database (2) 13 The DBS only acts as a storage for information. This information must first be entered and then kept current. Keeping a DBS current Ann Smith has done Homework 3 and received 10 points for it Database System ok. A DBS is computerized version of a card index box/filing cabinet (but more powerful and efficient). Task of a Database (3) 14 Normal database systems do not perform particularly complicated computations on the stored data in order to answer questions. However, a DBS can retrieve the requested data quickly from a huge set of data (giga bytes, tera bytes, main memory size). A DBS can also aggregate/combine several pieces of stored data to answer more complex questions ( Compute the average points for Homework 3. )

8 Task of a Database (4) 15 Above, the question Which homeworks has Ann Smith completed? was shown in natural (English) language. Making machines understand natural language is a tough task (and bears a large potential for misunderstandings). Therefore, questions (or queries) are normally written in a formal language, these days typically SQL. SQL SQL Structured Query Language, development started in 1986, current version SQL:2008. Pronounced S Q L, or Sequel. State, Query, Update 16 The set of stored data is called the database state: Current State Query SELECT HOMEWORK FROM SOLVED WHERE STUDENT = Ann Smith Answer Entering, modifying, or deleting information changes the database state: Current State Update INSERT INTO SOLVED VALUES ( Ann Smith, 3, 10) New State

9 Structured Information (1) 17 Each database can store only information of a predeclared structure (a limited domain of discourse): Structure mismatch Today s special in the cafeteria is pizza. Homework DBS Error. Because the data are structured, not simply text, complex query formulations are possible, e.g. How many homeworks has each student done? Structured Information (2) 18 Actually, a database system stores only plain data (character strings, numbers), and not information. Data becomes information by interpretation. Therefore, real world concepts like students, homework, cafeterias, etc., need to be defined/declared before the database can be used. A pure text database? Assume a DBS is storing text (character strings) only, with no further structure provided. What types of questions would you be able to ask?

10 State vs. Schema (1) 19 Database Schema: Formal definition of the structure of the database contents. Determines the possible database states. Defined only once (when the DB is created). In a programming language, this corresponds to variable declaration (assigning a type to a variable). Variable declaration Example: variable declaration in C: short int i Possible states of variable i? i State vs. Schema (2) 20 Database State (Instance of the Schema): Contains the actual data, structured according to the schema. Changes often (whenever database information is updated). Corresponds to current contents/value of a programming language variable. Variable state change In state s, variable i has value 41. Now perform state change (s to s ) via assignment i = i + 1.

11 State vs. Schema (3) 21 In the relational model, the data is structured in form of tables (relations). Each table has a name, a sequence of named columns (attributes), and a set of rows (tuples). A table SOLVED } DB Schema STUDENT HOMEWORK POINTS Ann Smith 1 10 } Ann Smith 2 8 DB State Michael Jones 1 9 Michael Jones 2 9 Data Model (1) 22 Defines a formal language (syntax & semantics) for declaring database schema querying the current database state changing the database state. 1 Examples: (Network Model, Hierarchical Model), Relational Model, Entity Relationship Model, Object Oriented Models, UML, XML. 1 Data model is, regrettably, widely used for Database schema.

12 Introduction 23 Overview 1. Basic Database Notions 2. Database Management Systems (DBMS) 3. Programmer s View, Data Independence 4. Database Users and Database Tools DBMS (1) 24 A Database Management System (DBMS) is an application independent software system that implements a data model, i.e., allows for definition of a DB schema for some concrete application, storage of an instance of this schema on, e.g., a disk, querying the current instance (database state), changing the database state. Application independent vs. concrete application Since a DBMS is application independent, how will the DBMS ensure to interpret the stored application data correctly?

13 DBMS (2) 25 Normal users do not need to use SQL for their daily tasks of data entry or data lookup. These users use application programs that have been developed specifically for this task and offer a more accessible user interface. Internally, these application programs translate the user requests into SQL statements (queries, updates) in order to communicate with the DBMS. DBMS (3) 26 Often, several different application programs are used to access the same centralized database. For example, the Homework DBS might provide: A read only web interface for students. A program used by the TA (Hiwi) to load homework and exam points. A program that prints a report for the professor used to assign grades. The interactive SQL interface (SQL console) that comes with the DBMS is simply yet another way to access the DBMS.

14 DBMS (4) 27 User A Application Program User B DBMS Tool (e.g., SQL console) Database Management System (DBMS) DB Schema DB State DB Application Systems (1) 28 Often, different users access the same database concurrently (i.e., at the same time, touching the same data). The DBMS is usually implemented as a background server process (or set of such processes) that is accessed over the network by application programs (clients). One can also view the DBMS as an extension of the operating system (a more powerful file system).

15 DB Application Systems (2) 29 Client Server Architecture Client Client Server Network User A (Application) User B (SQL console) DBMS DB Application Systems (3) 30 Three-Tier Architecture Thin client Thin client Application Server Server User A (Browser) User B (Browser) Application Web Server DBMS

16 DB Application System (4) 31 A recap of database vocabulary: A database (DB) consists of a DB schema and a DB state. A database management system (DBMS) is a software system that implements a data model (e.g., a Relational DBMS (RDBMS) implements the relational model). A database system (DBS) consists of a DBMS and a database. A database application system consists of a DBS and a set of application programs. Introduction 32 Overview 1. Basic Database Notions 2. Database Management Systems (DBMS) 3. Programmer s View, Data Independence 4. Database Users and Database Tools

17 Persistent Storage (1) 33 Today: 5 factorial 120 Tomorrow: 5 factorial 120 To evaluate factorial (n n!), no persistent storage is necessary. The output is a function of the input only. Persistent Storage (2) 34 Today: Ann Homework points 20 Tomorrow: Ann Homework points 30 The output is a function of the input and a persistent state.

18 Persistent Storage (3) 35 A DBS provides persistent state Input Ann Homework points Output 30 Persistent state Persistent information Information that lives longer than a single process (program execution). Survives power outage and a reboot of the operating system. Persistent Storage (4) 36 Which of the following processes/devices need persistent storage? If so, for which particular task? 1 Web browser 2 Pocket calculator 3 Mobile phone 4 Screen saver 5 HDD/DVD video recorder

19 Typed Persistent Data (1) 37 Classical way to implement persistence: Information needed in subsequent program invocations is saved into a file. The operating system (OS) maintains the file on disk. Disks provide persistent memory: the contents is not lost if the machine is switched off or the OS is rebooted. OS files and persistence The above statement is basically true but care should be taken nevertheless. Why? File systems are predecessors of modern DBMS. Typed Persistent Data (2) 38 Implementing persistence with files: OS files are usually nothing but sequences of bytes. A record structure must be defined on top of this (much like in Assembler languages): A n n S m i t h The record and file structure is contained only in the programmers heads. The OS file system cannot prevent misinterpretation, overflows, etc., because it is not aware of the file structure

20 Typed Persistent Data (3) 39 Implementing Persistence with a DBMS: The structure of the information to be stored must be defined in a way the DBMS understands: SQL DDL command CREATE TABLE SOLVED (STUDENT VARCHAR(40), HOMEWORK NUMERIC(2), POINTS NUMERIC(2)) The file structure is formally documented. The system can detect type errors in application programs. Simplified programming (higher abstraction level). A Subprogram Library (1) 40 Most DBMSs use OS files to store the data. (Some use raw disk device access.) One can view a DBMS as a subprogram library that can be used for file access. Compared with the direct OS system calls for file access, the DBMS offers higher level operations. The DBMS offers a wide variety of algorithms that one would otherwise have to program.

21 A Subprogram Library (2) 41 For instance, a typical Relational DBMS contains routines for sorting (e.g., external merge sort), searching (e.g., B-trees), file space management, buffer management, aggregation, statistical evaluation. The algorithms are optimized for large data sets (that do not fit into main memory). The DBMS also offers multi-user support (locking) and safety measures to protect data against system crashes. Data Independence (1) 42 The DBMS is a layer of software above the OS files. The files can be accessed only via the DBMS. The DBMS may change the file structure internally (reorder records, splits files, etc.) for performance reasons. This goes unnoticed by the application program. Compare with the idea of abstract data types: The implementation changes, the interface is kept stable.

22 Data Independence (2) 43 Typical example: At the beginning, a professor used the homeworks DB only for his courses in the current term. Since the DB was small and there were relatively few accesses, it was sufficient to store the data as a heap file. Later, the entire university used the DB, and information of previous courses had to be kept for some time. DB size grows significantly, DB access much more frequently. An index file (e.g., a B-tree) is now needed to provide fast access. Data Independence (3) 44 Without DBMS: Using the new B-tree index to access the file must be explicitly built into the lookup (query) commands. Thus, application programs need to be changed if the mode of file access is changed. If one forgets to change a seldolmly used application program, and this program does not update the index when the data has been updated, the DB becomes inconsistent.

23 Data Independence (4) 45 With Relational DBMS: Already at the interface, the system completely hides the (non-)existence of indexes on files. Queries and updates do not have to and cannot refer to indexes. The system automatically 1 modifies the index in case of data updates, 2 uses the index to evaluate queries against the indexed data when advantageous. Data Independence (5) 46 Conceptual Schema ( interface ): Only logical information content of the database, relevant to the subset of the real world modelled in the DB. Simplified view of the DB: physical storage details hidden. Internal/Physical Schema ( implementation ): Indexes, Division of tables among disks, Storage management if tables grow or shrink, Placement of new rows in a table (sort order, clustering).

24 Data Independence (6) 47 1 The user enters a query (e.g., in SQL) that only refers to the conceptual schema. 2 The DBMS translates this into a query/program (execution plan) which refers to the internal schema. This is done by the query optimizer. 3 The DBMS executes the translated query on the persistent instance of the internal schema. 4 The DBMS translates the result back to the conceptual level. Back-translation? Why would this be necessary and what would be typical back-translation steps? Data Independence (7) 48 Changing the internal schema Conceptual Schema Same Conceptual Schema New Translation Old Internal Schema (no B-tree index) New Internal Schema (with B-tree index)

25 Declarative Languages (1) 49 Physical data independence requires that the query language (SQL) cannot refer to indexes. Declarative query languages go one step further: Queries should only describe what information is sought, but should not prescribe any particular method how to compute/retrieve the desired information. Kowalski Algorithm = Logic + Control Imperative/Procedural Languages: explicit control, implicit logic Declarative/Descriptive Laguages: implicit control, explicit logic Declarative Languages (2) 50 SQL is a declarative language. The user describes conditions the requested data is required to fulfill: SQL query SELECT X.POINTS FROM SOLVED X WHERE X.STUDENT = Ann Smith AND X.HOMEWORK = 3 Ofter, simpler formulations of the same query are possible, with SQL users do not have to think about efficient execution. More concise than imperative programming: less expensive program development and maintenance.

26 Declarative Languages (3) 51 Declarative query languages allow powerful optimizers (no evaluation method is prescribed) need powerful optimizers (naïve evaluation is almost always too inefficient). Independence of current hardware technology and software quality: Today s queries will use tomorrow s DBMS setup and algorithms when a new version of the DBMS is released. Logical Data Independence (1) 52 Logical data independence allows for changes to the logical information content of the database. Such changes are obviously restricted to additions to the logical information content. Example: add column SUBMISSION DATE to table SOLVED. Such additions may be required for new applications. It should not be necessary to change old applications only because records now contain additional information.

27 Logical Data Independence (2) 53 Logical data independence is important when there are application programs with distinct, but overlapping information needs. Logical data independence also helps to integrate previously distinct databases. In earlier times, every department of a company had its own DB/data files. Now, businesses generally aim at one central DB. Logical Data Independence (3) 54 If a company uses more than one DB, the information in these databases will normally overlap, i.e., some pieces of information will be stored several times. Data is called redundant if it can be derived from other data and knowledge internal to the application. Problems with redundancy: Duplicates data entry and update efforts. Sooner or later, data copies will get out-of-sync and thus inconsistent. Wastes storage space, also on backup media.

28 Logical Data Independence (4) 55 External Schemas/Views: Logical data independence requires a third level of database schemas, the external schemas or views. Each user (department,... ) may have an individual view of the data. An external view contains a subset of the information in the database, maybe slightly restructured. Views may also be vital because of security reasons. In contrast, the conceptual schema describes the complete information content of the database. Three Schema Architecture 56 Three Schema Architecture [ANSI/Sparc 1978] User User External Schema 1... External Schema n Conceptual Schema Stored data Internal Schema

29 More DBMS Functions (1) 57 Transactions: Sequences of DB commands (queries and updates) are executed as an atomic unit ( all or nothing ). DBMS may crash during/after a sequence of commands is/has been executed. The DBMS then performs undo/redo. Support for backup and recovery. Support of concurrent users. Each user is given the illusion to be the only DB user at any time. DBMS performs locking and conflict detection. More DBMS Functions (2) 58 Security: Access rights: Who may perform which operations on which table? Auditing: DBMS remembers who did what/when. Integrity: The DBMS checks that the entered data is plausible/complete (such checks may also span several tables). DBMS rejects updates (insertions and deletions) which would violate defined business rules.

30 More DBMS Functions (3) 59 Data Dictionary: Metadata ( data about data, schema, user list, access rights) is availble in system tables, e.g.: System tables SYS TABLES TABLE NAME OWNER SOLVED GRUST SYS TABLES SYS SYS COLUMNS SYS SYS COLUMNS TABLE NAME SEQ COL NAME SOLVED 1 STUDENT SOLVED 2 HOMEWORK SOLVED 3 POINTS SYS TABLES 1 TABLE NAME SYS TABLES 2 OWNER SYS COLUMNS 1 TABLE NAME SYS COLUMNS 2 SEQ SYS COLUMNS 3 COL NAME Introduction 60 Overview 1. Basic Database Notions 2. Database Management Systems (DBMS) 3. Programmer s View, Data Independence 4. Database Users and Database Tools

31 Database Users (1) 61 Database Administrator (DBA): Should know about all schemas, may change the conceptual and the internal schema (creates tables, creates/drops indexes). Can damage everything. Gives access rights to users. Ensures security. Monitors system performance. (Transaction throughput #TX /s, # concurrent users, index sizes,... ) Monitors available disk space and installs new disks. Ensures that backup copies of the data are made. Does recovery after disk failures, etc. Database Users (2) 62 Application Programmer: Writes programs for standard, all-day tasks, to be used by the naïve users (see below): safe data entry, report generation, data browsing. Knows SQL well, plus programming languages and development tools. Usually supervised by DBA. Might do conceptual schema design (knows which table the application will need to access/create).

32 Database Users (3) 63 Sophisticated User (one kind of end user ): Knows SQL and/or some query tools, may use SQL console. Does non-standard aggregations/evaluations of the data without help from application programmers. May generate complex queries. Naïve User (the other kind of end user ): Uses DB only via application programs, often unaware of existence of DBMS back-end. Primarily data entry user, simple browsing-style queries against external views. Database Tools 64 Interactive SQL console Graphical/menu-based query tools Interface for DB access from standard programing lanugages (C, C++, Java) Tools for form-based DB application (4GL) Report generators Web interface Tools for data import/export, backup & recovery, performance monitoring,...

33 Summary (1) 65 Functions of database systems: Persistence Integration/Redundancy Avoidance Physical and Logical Data Independence Subprogram Library: many algorithms built-in, especially tuned for external memory access (disks) Query and Update evaluation Summary (2) 66 Functions of database systems, continued: High data safety and availability (Backup & Recovery) Combinations of operations into atomic transactions Multi-user support: synchronization of concurrent accesses Integrity Enforcement View management Security via data access control System catalog management (metadata)

34 Summary (3) 67 The main goal of the DBMS is to give the user a simplified view on the persistent storage, i.e., to hide any complications introduced by the DBMS physical layer. The user does not worry about physical storage details different information needs of other users efficient query formulation possibility of system crashes/disk failures presence of concurrent users accessing identical data subsets. Exercise 68 Data-intensive programming Suppose homework points data is stored in a line-structured text file with the format Student Name:Homework Number:Points e.g., Ann Smith:3:10 Suppose you have to write a C program that prints the total number of points per student (students sorted alphabetically). How would you judge the programming effort (in terms of lines and time)? In SQL, this takes 4 lines and approx. 1 1 /2 minutes.