DATABASES AND THE INTERNET - Fundamentals of Database Management Systems - page 375

of the Internet and Web environment provides many potential reasons for poor
performance, including whether your connection to the Internet is through a 56K
modem or a broadband connection, the level of hardware at your Internet service
provider, the speed of the Web server you are interacting with, associated facilities
at the Web site, and so forth.
From the point of view of the company and its Web site, there is another major
factor, too: the amount of traffic coming in from the Internet. Internet traffic to a
Web site, the number of people or companies trying to access it simultaneously, can
vary greatly because of a variety of factors:
■ The time of day (which must be considered on a worldwide basis).
■ The season of the year (e.g. the Christmas shopping season).
■ The rapidly growing popularity of a Web site.
■ A major new product introduction.
■ A major event (e.g. the Victoria's Secret annual fashion show, which has
overwhelmed its Web site).
These spikes, some of them huge, in Internet traffic require serious predictive
capacity planning. The trick is that the companies want to be able to maintain
reasonable response time during the spikes without spending lots of money to buy
lots of extra computer equipment that will sit idle much of the time. Accomplishing
this takes some serious planning and significant expertise.
Of course, system performance is also affected by software design and, in
particular for our interest here, in database design. Thus, Chapter 8 on physical
database design is of particular interest in the Internet database environment.
The various performance-boosting physical design techniques that we discussed,
including denormalization, are applicable at least for relatively static database tables
such as product lists in some industries.
In addition, we mention two performance-boosting techniques that are of
particular interest in the Web database environment. For the first one, take another
look at Figure 14.4. When a query comes in from a PC and is passed from the Web
server to the database server to the database, it is often the case that the retrieved
data will be used again by the same or a different end user fairly soon. If a copy
of that data can be held somewhere outside the database on a temporary basis for
the next time it is requested, then two benefits can be gained: not only can the
response time for future retrievals of that data be improved, but the amount of
traffic between the Web server and the database server can be decreased, which
helps to improve the performance of other accesses to the database. This concept of
database persistence can be accomplished with a query cache , a special dedicated
memory associated with the Web server or a proxy server attached to it, to hold a
copy of the retrieved data temporarily. A second performance-boosting technique is
used in situations where company employees can write SQL queries to access data
over the Internet (or over an internal Intranet): frequently run queries can be stored
or ''canned'' and then called when needed. This avoids having the system spend
time going through query optimization to come up with an efficient access path
every time the query is run, a concept that was discussed in general in Chapter 4.
Availability
A company's Web site and the databases it accesses should be available to the
public at all times. This is especially true if the company is expecting traffic to the