databases notions slide 11
play

Databases notions - PowerPoint PPT Presentation

Databases notions slide 11 Technical context DB advantages DB vocabulary DB data models slide 31 A Reference


  1. ������������ ������������������������� � Databases notions slide 11 � Technical context � DB advantages � DB vocabulary � DB data models slide 31 � A Reference architecture slide 44 � What is a DBMS ? slide 49 10 �������������������� ������������������� � Principles � Data in main-memory (variables) � Data storage in a fast access but non persistent � Files storage � Databases � Problems � Temporary storage � «low-capacity » � Programming languages � Single-process (single-user) access 11 12 Introduction- 1

  2. ���������������������� ����� � Persistent storage on disks Study direction library Social service � «high-capacity» of data Cobol Java appli C Appli Appli � Programming languages and file systems � Single-process (single-user) access Students and Modules files Students and Books files Students and Rooms files 13 14 ������������������ ������������������� � No global vision of data Library application � Problem to understand links between data Program 1 � Data redundancy Description of Students file Edition of Students file in alphabetical � Inconsistency order * 2 � No data sharing between users Program 2 � No independence between data and programs Description of Sudents files Students file Updates on Students giving a name � Problems with data security FS � Multiplicity of languages, systems, hardware 15 16 Introduction- 2

  3. ��������� �������� ����������� � Persistent storage on disks PC Windows DB � «very high-capacity» C Appli Students Modules � Query languages and database PLs Cobol Appli Java Appli Books Rooms � Multi-users PC DBMS NT DB catalog PC Linux Terminals 17 18 ������!���"����������� ������������"����#�� � Integration : PC Windows DB � Unique and global description of data Students Cobol Appli Modules � No redundancy Books � Consistency Rooms DBMS PC � Independence : Java Appli NT � Independence between data and programs DB catalog � Independence between a logical and physical description of C Appli data PC Linux 19 20 Introduction- 3

  4. ������������"����#���$%& ����������� � Data � Security : � Data model � Semantic control � Database � Protection against unauthorized access � Protection against crashes � DBMS � Easiness for end-user : � Database schema � Data sharing � Database instance � Personalized and High-level vision � Easy management of data � Integrity constraint � Efficient access to data � Data dictionary (catalog) � Distribution of data and programs 21 22 ����� �����$%& � A structure � Users � Simple: price, name, date � Querying � Complex: person, document, image � A semantic : � « temperature on Paris the 6th of October 2004 ? » � Amount of money on an account � Updating � A picture of Paris � « Add 100 euros on M. Hill’s account » � An owner : � Who creates the data � Who defines rules on the data « the temperature has to be between -30° C and +40° C » � « the salary of year n has to be greater than the salary of year n-1 » � � Who gives rights on the data 23 24 Introduction- 4

  5. ���������� ����������������� � DB : � Set of concepts to describe : � Collection of data described according to a model � Data of the real world � DBMS : � Links between data � Semantic of data � Software managing data, according to a model � Set of operators to manage data � A DBMS has to allow data definition, data updates and data control 25 26 ���������������������� ���������������������� � Schema (intension) : � Integrity constraint (IC) : � Data description, according to a model � Rule defined on data, to specify a consistent state of � Network model, relational model, hierarchical model the database � Generally static � The salary has to be greater than the “1000” � Instance (extension) : � Data dictionary (DD): � Data Collection, described according to a model � data describing the data (meta-data) � Instance of the schema � « values of the schema » � Dynamic 27 28 Introduction- 5

  6. '��������������� �������������� � As a central element : Data volume Program Service intensity � databank (professional, general public) complexity � Transactional systems (bank, insurances,…) � Decision support system (dashboard, statistical analysis,…) Transactional ≈ Go Querying + > 100 tps � As an essential element : application Updates � Communication and Information systems (enterprise, administration,…) Response time < 2 s. � Control and supervision Systems (telecommunications, production,…) � As an auxiliary element : Decision support ≈ To Querying Tpm notion � Information retrieval in digital libraries systems (texts, sounds, graphics, system Response time …) Datewarehouse in mn or h � CAx systems (CAD, CALS, …) Datamining � … TPC (Transaction Processing Council) Benchmarks transactional systems 29 30 �������������#� �������������������������� � Data design Reality � Models Very high-level � Entity relationship model [Chen 76] ER Model, SDM, Z Conceptual schema � Hierarchical model Taking into account DBMS features � Network model High-level Network or relational models Conceptual schema DDL � Relational model Internal schema Tools: placing and access methods 31 32 Introduction- 6

  7. (������������������������� ���������#� )�����*+, moduleId nbH coord Data model : � Set of concepts to describe : � Module Data � Links between data � 1,n Semantic of data studentId � mark Register Generally, an associated set of operators � Description formalism : � 3,n name Textual � Student Graphical � age 0,n 0,1 Mathematics � Examples of models : � address Entity-relationship � Hierarchical / Network model � Relational � lendingDate Borrow Rent Object, Relational-Object � 0,1 bookId 1,1 roomId Book Room title price 33 34 (��������������������-% '�"����#��!�������������(.������ � Advantages � � drawbacks � Rich semantic � Just for data describing Book � Extension to object � No associated operators 0,1 bookId Borrow concepts (inheritance, ...) title � No ER DBMS Student lendingDate Module 0,n studentId � visual Register moduleId 1,n 3,n name nbH mark � model for database � � not an implementation age Rent 0,1 coord address Room design model roomId 1,1 price 35 36 Introduction- 7

  8. /����������������� /������������������$%& � DB Schema � IBM IMS system, designed at the end of 60ies for � Tree structure Appolo program (NASA) � DB � Example � Set of records linked by pointers � DML � Navigational and procedural language Students Modules � (DL/1, IMS system) � Problems : � No independence between logical/ physical level � Data Redundancy => inconsistency Books Modules Rooms Students 37 38 0������������ 0�������������$%& � DB schema � Defined by the DBTG, CODASYL comity, 1971 � Acyclic oriented graph � DB (new version in 1978) � Set of records linked by pointers � Example � DML � Navigational and procedural � Pointer based Students Modules � Standards CODASYL 71, 78 � Systems � IDS2 of Honeywell (1975), Total of Cincom (1974), Adabas of Soft. Ag (1978) � Problems Book Room Registration � No physical/logical independence 39 40 Introduction- 8

Recommend


More recommend