Principles of Database Management Systems CSE 544 Introduction March 31st, 1999
Staff Instructor: Alon Levy TAs: Zack Ives and Rachel Pottinger Sieg, Room 310, alon@cs.washington.edu Office hours: wed, 2:30-3:30. Or by email. TAs: Zack Ives and Rachel Pottinger Office hours: Zack: Mondays at noon (224) Rachel: Thursdays at 2:30pm (224) Mailing list: cse544@cs Web page: (a lot of stuff already there) http://www.cs.washington.edu/education/courses/544/99sp/
Course Times In general, WF, 12-1:20pm (with a 5 minute breather in the middle). Two special dates: Monday, April 5th Monday, April 19th No classes on last week.
Goals of the Course Purpose: Foundations of database management systems. Issues in building database systems. Introduction to current research issues in databases. Have fun: databases are not just bunches of tuples.
Grading Homeworks: 15% Project: 25% Midterm: 15% Final: 35% SQL querying fun Join implementations Project: 25% A query optimization engine for data integration. Midterm: 15% Final: 35% Participation and intangibles: 10%
Textbook Database System Implementation, Ullman, Widom, and Garcia-Molina, to be published by Prentice-Hall in June; available from the copy center.
Other Useful Texts Database Management Systems (Ramakrishnan) Foundations of Databases (Abiteboul, Hull & Vianu) Parallel and Distributed DBMS (Ozsu and Valduriez) Transaction Processing (Gray and Reuter) Database Systems (Silberschatz, Korth and Sudarshan) Data and Knowledge based Systems (volumes I, II) (Ullman) Readings in Database Systems (Stonebraker and Hellerstein) Proceedings of SIGMOD, VLDB, PODS conferences.
Prerequisites
Real Prerequisites Operating systems Data structures and algorithms Distributed systems Complexity theory Mathematical Logic Knowledge Representation User interface design Programming languages Artificial Intelligence (Search) Greek, Hebrew, French
Why Use a DBMS? All programs manipulate data, so why use a database? Large amounts of data (Giga’s, Tera’s) Data is very structured Persistent data Valuable data Performance requirements Concurrent access to the data Restricted access to data
Functionality of a DBMS Persistent storage management Transaction management Resiliency: recovery from crashes. Separation between logical and physical views of the data. High level query and data manipulation language. Efficient query processing Interface with programming languages
Persistent Storage Becomes a hard problem because of the interaction with the other levels of the DBMS: What are we storing? Efficient indexing Special issues due to resiliency requirements Exploit “semantic” knowledge Issue: interaction with the operating system. Should we rely on the OS?
Transaction Processing and Recovery For efficient use of resources, we want concurrent access to data. Systems sometimes crash. A “real” database guarantees ACID: Atomicity: all or nothing of a transaction. Consistency: always leave the DB consistent. Isolation: every transaction runs as if it’s the only one in the system. Durability: if committed, we really mean it. Do we really want ACID?
Physical vs. Logical Levels External Schema 1 External Schema 2 Conceptual schema: tables and their attributes Physical schema: files, indexes hash tables. External schema: views of the different applications, classes of users. Relational Schema System catalog: The component of the database that stores meta data. Conceptual design: a precursor to the relational schema. Physical Schema Disk
The Relational Model Data is organized into tables with attributes. Rows in the tables are tuples. The power of simplicity!
Logical Model Issues What data model should we use? Relational, object-oriented, object-relational, deductive database model, semi-structured How do we design a good schema? (normal forms, index selection) Are we really providing an abstraction? How does this abstraction interact with the programming language? (the impedance mismatch).
Querying a Database Find all the students who have taken CSE444 in Winter, 1998. S(tructured) Q(uery) L(anguage) select E.name from Enroll E where E.course=CSE444 and E.quarter=“Winter, 1998” SQL also provides update facilities. SQL: an acquired taste (try datalog first)
Issues in Query Languages Does it provide the appropriate functionality? SQL books get thicker and thicker. Expressive power of a query language. Ease of use (query by example) Declarativity Provide guidance in writing “good” queries?
Query Optimization A query is a declarative specification of “what” you want. A query execution plan is an imperative program to produce the answer. Query optimization: produce an efficient query execution plan. Issues: large search space of plans, cost estimation, semantic transformations Real goal: avoid the bad plans.
Database Industry Relational databases are a great success of theoretical ideas. “Big 3” DBMS companies are among the largest software companies in the world. IBM (with DB2) and Microsoft (SQL Server, Microsoft Access) are also important players. $20B industry Moving to warehousing, decision support.
Course (Rough) Outline The basics: The relational model SQL Views, integrity constraints Conceptual modeling datalog (recursive queries) Physical representation: Index structures.
Course Outline (cont) Query execution: Query Optimization Algorithms for joins, selections, projections. Query Optimization Advanced topics: data integration data mining semi-structured data Transaction processing
The relational data model
Terminology Product (relation name) Attribute names Name Price Category Manufacturer gizmo $19.99 gadgets GizmoWorks Power gizmo $29.99 gadgets GizmoWorks SingleTouch $149.99 photography Canon MultiTouch $203.99 household Hitachi tuples (Arity=4) Product(name: string, Price: real, category: enum, Manufacturer: string)
More Terminology Every attribute has an atomic type. Relation Schema: relation name + attribute names + attribute types Relation instance: a set of tuples. Only one copy of any tuple! (not) Database Schema: a set of relation schemas. Database instance: a relation instance for every relation in the schema.
More on Tuples Formally, a mapping from attribute names to (correctly typed) values: name gizmo price $19.99 category gadgets manufacturer GizmoWorks Sometimes we refer to a tuple by itself: (note order of attributes) (gizmo, $19.99, gadgets, GizmoWorks) or Product (gizmo, $19.99, gadgets, GizmoWorks).
Integrity Constraints An important functionality of a DBMS is to enable the specification of integrity constraints and to enforce them. Knowledge of integrity constraints is also useful for query optimization. Examples of constraints: keys, superkeys foreign keys domain constraints, tuple constraints. Functional dependencies, multivalued dependencies.
Keys A minimal set of attributes that uniquely the tuple (I.e., there is no pair of tuples with the same values for the key attributes): Person: social security number name name + address name + address + age Perfect keys are often hard to find, but organizations usually invent something anyway. Superkey: a set of attributes that contains a key. A relation may have multiple keys: (but only one primary key) employee number, social-security number
Foreign Key Constraints Purchase: buyer price product Joe $20 gizmo Jack $20 E-gizmo Product: name manufacturer description gizmo G-sym great stuff E-gizmo G-sym even better An attribute of a relation R is must refer to a key of a relation S.
Functional Dependencies Definition: If two tuples agree on the attributes A , A , … A 1 2 n then they must also agree on the attributes B , B , … B 1 2 m Formally: A , A , … A B , B , … B 1 2 n 1 2 m Key of a relation: all the attributes are either on the left or right.
Some Obvious Properties of FD’s A , A , … A B , B , … B Is equivalent to 1 2 n 1 2 m A , A , … A B 1 2 n 1 Splitting rule and Combing rule A , A , … A B 1 2 n 2 … A , A , … A B 1 2 n m A , A , … A A Always holds. 1 2 n i
Comparing Functional Dependencies Entailment: a set of functional dependencies S1 entails a set S2 if: any database that satisfies S1 much also satisfy S2. Example: A B, B C entails A C Equivalence: two sets of FD’s are equivalent if each entails the other. {A B, B C } is equivalent to {A B, A C, B C} Closure: Given a set of attributes A and a set of dependencies C, we want to find all the other attributes that are functionally determined by A.
Closure Algorithm Start with Closure=A. Until closure doesn’t change do: if is in C, and B is not in Closure then add B to closure. A , A , … A B 1 2 n A , A , … A Are all in the closure, and 1 2 n
Problems in Designing Schema Name SSN Phone Number Fred 123-321-99 (201) 555-1234 Fred 123-321-99 (206) 572-4312 Joe 909-438-44 (908) 464-0028 Joe 909-438-44 (212) 555-4000 Problems: - redundancy - update anomalies - deletion anomalies
Relation Decomposition Break the relation into two relations: Name SSN Fred 123-321-99 Joe 909-438-44 Name Phone Number Fred (201) 555-1234 Fred (206) 572-4312 Joe (908) 464-0028 Joe (212) 555-4000
Boyce-Codd Normal Form A simple condition for removing anomalies from relations: A relation R is in BCNF if and only if: Whenever there is a nontrivial dependency for R , it is the case that { } is a super-key for R. A , A , … A B 1 2 n A , A , … A 1 2 n In English (though a bit vague): Whenever a set of attributes of R is determining another attribute, should determine all the attributes of R.