DISTRIBUTED DATABASE MANAGEMENT SYSTEM

1
DISTRIBUTED DATABASE MANAGEMENT SYSTEM

• Azeem Khan
• Tony Su
• Peter Lymberatos
• Yi Liang
• CUS 1165
• Spring 2001
• Prof. Boehrer

2

• Introduction Definition
• Architecture
• Partitioning
• Replication
• Class participation

Overview
3
Introduction Definition
Managing multi-user databases   Database
applications on the Internet or on organizational
intranets can have dozens, hundreds, or even
thousands of users. The databases in such systems
become immeasurable valuable assets.   History
of Managing Multi-user Databases
Mainframe database processing ? client-server
database processing ? Internet and intranet
database processing  
4
Introduction Definition
Four important topics regarding multi-user
database management. 1) Concurrency control 2)
Database reliability 3) Database security 4)
Database administration -- Managing the
database structure  -- Managing data
activity -- Managing the DBMS --
Maintaining the data repository


5
Introduction Definition
( Distributed database management systems
DDBMSs)   With a distributed database
system, multiple computers process the same
database. The basis premise behind a distributed
database system is to make a data store spread
over a collection of machines in a wide or local
area network appear to the user as if it existed
on a single machine.
6
Introduction Definition

• There are several types of distributed databases
  
• partitioned -- nonreplicated
• non partitioned -- replicated
• partitioned -- replicated.

7
Introduction Definition

• DDBMS advantages
• data relocated near the "greatest demand" site
• faster data access locally than remotely located
  
• centralized data
• faster data processing by spreading out the work
  load
• growth facilitation as it enables adding new
  sites
• without affecting other sites
• improve communications as sites are more closely
  
• located to customers
• reduced operating costs as less costly than to
  update a
• mainframe
• user friendly interface
• less danger of a single-point failure
• processor independence as any processor can
  access
• data at different location

8
Introduction Definition

• DDBMS disadvantages
• complexity of management and control
• security is less sophisticated than mainframe
• lack of standards as it is still not mature

9
Introduction Definition
The desired function of a distributed DBMS
can be explained by a collection of
transparencies -- Location transparency
-- Performance transparency -- Copy
transparency -- Transaction transparency
-- Fragment transparency -- Schema change
transparency -- Local DBMS transparency
10
Introduction Definition

• Application
• Microsoft -- SQL
• Oracle -- 8!
• Sybase -- Sybase's flagship Adaptive Server
  Enterprise
• Informix -- Informix Dynamic Server
• IBM -- DB2 Universal Database

11
Introduction Definition

• Distributed Database Cases
• Sea-Land Service, Inc.
• E-Plus Mobilfunk GmbH
• Surridge Dawson Ltd.
• Northwest Airlines

12
Architecture
Three types of architecture
(1) Star (2) Ring (3)
Mesh
13
(1) Star
All the link of data are centralized
in one area called HUB and then distribute
to the end users through their links.
14
(1) Star
15
(1) Star
A) Advantage -- With a hub, if one
of the link fails the remaining links
will still work fine. -- It is very easy to
manage and monitor the data. B) The
disadvantage -- If the hub breaks down, all
the data may be destroyed and all the links will
be broken. -- The cost. Since large company
need to sent data to many different locations,
the links can be a very big part of expense.
16
(2) Ring
In a Ring distribution, all the data are
connected together in a loop format, that is that
the data loops around all the locations. It has
no start and end.
17
(2) Ring
18
(2) Ring

• Advantage
• -- All the end users in each location has equal
  access to communicate or exchange the data with
  each other.
• -- If one location sends out lot of data, it
  wont slow down the entire database.
• Disadvantage
• -- If one connection of the loop fails, the
  entire database will fail.
• -- Is very hard to find the cause of the
  problem.

19
(3) Mesh
The data stored in each location can be
shared by all end users, because simply all the
locations are fully connected. Every location
have a link to all other locations. Under
certain circumstances sometimes this kind of
distribution can be partially connected, which
means one or more of the locations are connected
with some other locations, but not all, only the
ends users can get the data by reroute the data
from other locations.
20
(3) Mesh
21
(3) Mesh

• Advantage
• -- Fault tolerance, if there is a break in the
  link , the data can be rerouted and is almost
  impossible for this kind of distribution to fail.
  
• B) Disadvantage
• With all the links connecting to all the
  locations is very hard to manage, and also very
  expensive.

22
Partitioning
Two types of partitioning (1)
Vertical partitioning (2) Horizontal
partitioning
23
(1) Vertical partitioning
Vertical partitioning is the subset of a table
consisting of complete rows of the table, which
takes attributes of a relation and groups them
together to create fragments.   These fragments
are then allocated, in a distributed database
system. There are allocated in such a way that it
helps increase performance of the system. The
partitioning process tries to match the fragments
to the transaction in order to reduce the
transaction processing costs. Typically the
partitioning algorithm deals with disjoint
fragments on non-primary key attributes. The
allocation of fragments allows parallel execution
of queries and reduces the amount of irrelevant
data accesses.
24
(2) Horizontal partitioning
Horizontal partitioning is a subset of a table
consisting of complete rows of a table. With
horizontal partitioning it uses a technique for
reducing the number of disk accesses to execute a
query by minimizing the number of irrelevant
instances accessed.   There are many advantages
of horizontal partitioning. Some of them are that
data is stored close to where it is used and also
there is local access optimization.
25
Replication

 

• What is replication?
• Two types of models
• Master/slave (Primary) replication
• Update anywhere replication

 
26
Class participation
Final question to the whole class Fill in the
blank please With a distributed database system,
multiple computers process the ______ database.
27
Bibliography

• Burleson, Donald K. (1994) Managing Distributed
  Databases.
• New York John Wiley Sons
• http//faculty.ed.umuc.edu/meinkej/inss690/stolli
  ngs/Stollings2069
• 020Final.htm
• http//lsewww.epfl.ch/Documents/html/WPS00b.html
  
• http//www.techweb.com/encyclopedia/defineterm.cgi
  ?sstringreplication
• Readings in Database Systems second edition
• Edited by Michael Stonebraker, University of
  California,
• Berkeley
• Object-Oriented Databases
• By EZ Nahouraii and Fred Petry
• Distributed Systems by Sape Mullender New
  York, New York Addison-Wesley Publishing
  Company , 1980
• Advances in Distributed and Parallel Knowledge
  Discovery Philip Chan MIT Press California
  1994