Enhanced Data Models for Advanced Database Applications

1
Enhanced Data Models for Advanced Database
Applications

• CS 95 Advanced Database Systems
• Handout 8

2
Conventional databases

• for large amounts of current only, complex,
  volatile, structured data, available within an
  organization
• e.g., Relational Databases
• Enhanced Data Models
• Active
• Temporal
• Spatial
• Multimedia
• Statistical
• Information retrieval

3
Active Databases

• contain a set of active rules
• Rule is triggered when a particular event occurs.
  eg. on an Update operation on a certain table
• The rule may initiate other, or even replacement
  operations to be performed on the database

4
Active Databases

• Usually implemented with triggers, a rule
  implementation found in many relational products.
• Such rules may be used to
• notify (someone) whenever a particular condition
  occurs
• create audit trails
• enforce integrity constraints
• maintenance of derived data
• maintain consistency of views whenever the base
  table structures are modified.

5
Active Databases

• Syntax summary for specifying triggers in the
  Oracle System
• lttriggergt CREATE TRIGGER lttrigger namegt
• (AFTERBEFORE) lttriggering eventsgt ON lttable
  namegt
• FOR EACH ROW
• WHEN ltconditiongt
• lttrigger actionsgt
• lttriggering eventsgt lttrigger eventgt OR
  lttrigger eventgt
• lttrigger eventgt INSERTDELETEUPDATE OF
  ltcolumn namegt, ltcolumn namegt
• lttrigger actiongt ltPL/SQL blockgt
• Even PostgreSQL supports triggers. See the
  PostgreSQL Help for more information.

6
Temporal databases

• conventional databases without the 'current only'
  constraint
• 'remembers' every value, i.e., when an update
  occurs, old value, new value and time of change
  all have to be stored in the DB.
• Different systems implement temporal data
  differently
• as a minimum a timestamp column is added to each
  relation and forms part of the key

7
Temporal databases

• Applications include
• Healthcare where the history of a patient is
  important
• Insurance
• Reservation systems
• Scientific databases
• SQL (usually) is extended so a query can provide
  an answer from the database at any specified
  time. eg What was the status of supplier S1 on
  April 1, 1990?

8
Temporal databases

• TSQL standard specifies the TIME PERIOD as a the
  standard domain measurement
• A query over a time PERIOD may return many rows
  for the same piece of information - one row for
  every set of values the information held within
  the time period.
• New SQL operations are like CONTAINS, PRECEDES,
  SUCCEEDS, OVERLAPS
• Problems DB grows very big very quickly, which
  adversely affects performance - dramatically.

9
Spatial databases

• conventional databases with spatial features,
  such as support for 3-dimensional objects
• applications
• CAD/CAM
• cartographic (mapping) - 2-D maps plus a
  description behind each object (road, bridge,
  house etc.)
• meteorological - weather patterns occur in 3-D
• drawing, drafting, etc.

10
Spatial databases

• Some objects may be static e.g., bridge, road
  etc. Other objects may have dynamic qualities
  e.g., ambulances and other vehicles.
• Even a simple task, such as the distance between
  two objects or determining whether two objects
  overlap, are difficult SQL queries so new
  operators are included to perform these tasks
• hence, spatial databases are often OODBMS

11
Spatial databases

• Types of queries particular to spatial database
  are
• Range query e.g., find all hospitals within a
  particular distance from a given location.
• Nearest neighbor query e.g., find an object of a
  given type closest to a given point.
• Spatial joins or overlays Joins the objects of
  two type. e.g., find all cities on a particular
  road OR find all homes near a given river.
• PostgreSQL supports many spatial objects,
  including point, line, box, circle etc.

12
Multimedia databases

• Multimedia databases store data such as images,
  audio and video clips and documents.
• Support content-based queries, e.g., retrieve all
  video clips with a certain person, or all clips
  containing a drag race won by a certain Pro Stock
  driver

13
Multimedia databases

• Two methods of identifying such content
• Automatic analysis - involves mathematical
  analysis and pattern matching of data within the
  clip. A different approach is required for video,
  audio, image and text. Obviously, this is a very
  computative intensive process.
• Manual identification - manual preprocessing
  phase obtains such information which is stored
  with the clips and can also be used to build
  indexes.

14
Statistical databases

• for large amounts of simple, non-volatile,
  structured data
• access is via aggregate statements only - SUM,
  COUNT, MAX, MIN, AVERAGE, STANDARD DEVIATION etc.
  
• they protect the identity of the individual
• problems with compromise of the DB.
• i.e., a set of queries that may determine the
  identity of an individual.
• Often solved by not allowing results that are
  computed from a return of less than n rows.

15
Information Retrieval databases

• Read-only, very large databases
• heavily indexed - indexes take a long time to
  update, so the data is not updated in real time.
  New data snapshots with their indexes are built
  periodically
• e.g., phone lists and search engines
• e.g., AltaVista indexes every word on every web
  page in its database
• index consists of word, location within page, URL
  
• size is ??? GB (200GB in 1998)