Database Reference
In-Depth Information
To illustrate the importance of minimizing the number of messages, suppose you have a network with
an access delay of 2 seconds and a transmission rate of 750,000 bits per second. Also suppose you send a
message that consists of 10,000 records, each of which is 800 bits long. (The 10,000 records is equivalent
to approximately 250 pages of single-spaced text.) In this example, you calculate the communication time
as follows:
Communication time
¼
2
þ
((10,000 * 800) / 750,000)
¼
2
þ
(8,000,000 / 750,000)
¼
2
þ
10.67
¼
12.67 seconds
If you send a message that is 100 bits long, your communication time calculation is as follows:
Communication time
¼
2
þ
(100 / 750,000)
¼
2
þ
.0001
¼
2.0001 seconds or, for practical purposes,
¼
2 seconds
279
As you can see, in short messages, the access delay becomes the dominant factor. Thus, in general, it is
preferable to send a small number of lengthy messages rather than a large number of short messages.
CHARACTERISTICS OF DISTRIBUTED DBMSs
Because a DDBMS effectively contains a local DBMS at each site, an important property of DDBMSs is that
they are either homogeneous or heterogeneous. A homogeneous DDBMS is one that has the same local
DBMS at each site. A heterogeneous DDBMS is one that does not; there are at least two sites at which the
local DBMSs are different. Heterogeneous DDBMSs are more complex than homogeneous DDBMSs and, con-
sequently, have more problems and are more difficult to manage.
All DDBMSs share several important characteristics. Among these characteristics are location transpar-
ency, replication transparency, and fragmentation transparency.
Location Transparency
The definition of a distributed database says nothing about the ease with which users access data that is
stored at other sites. Systems that support distributed databases should let a user access data at a remote
site
—
a site other than the one at which the user is located
—
just as easily as the user accesses data from the
local site
the site at which the user is located. Response times for accessing data stored at a remote site
might be much slower, but except for this difference, a user should feel as though the entire database is
stored at the local site. Location transparency is the characteristic of a DDBMS that users do not need to be
aware of the location of data in a distributed database.
—
Replication Transparency
As described in Chapter 7, replication lets users at different sites use and update copies of a database and
then share their updates with other users. However, data replication creates update problems that can lead to
data inconsistencies. If you update the record of a single part at Premiere Products, the DDBMS must make
the update at every location at which data concerning this part is stored. Not only do multiple updates make
the process more time-consuming and complicated, but should one of the copies of data for this part be over-
looked, the database would contain inconsistent data. Ideally, the DDBMS should correctly handle the updat-
ing of replicated data. The steps taken by the DDBMS to update the various copies of data should be done
behind the scenes; users should be unaware of the steps. This DDBMS characteristic is called replication
transparency.
