Databases Reference
In-Depth Information
The situation was unacceptable; the company had to find the performance
issues in its database application deployment. DataBank started by making sure
that its database was optimally tuned. Even with the database performing well,
the company was still missing its contractual service-level requirements.
The system architect made a phone call to a database consultant, and the
consultant asked, “Have you considered trying a different database driver?” The
architect responded, “I didn't even know that was an option.” The consultant rec-
ommended a database driver that he had used with success.
Losing no time, the architect had the recommended database driver installed
in a test environment. Within two days, the QA department reported a threefold
improvement in average response time between the new and the currently
deployed database drivers, as well as the elimination of stability problems.
Based on the results of its performance testing, DataBank moved forward to
purchase the new database driver. After the new database driver had been
deployed for a couple of months, DataBank analyzed the revenue it was saving.
DataBank was paying $250,000 in fines in September and reduced that to
$25,000 by November. That is a savings of 90% in two months by simply chang-
ing the database driver. The new driver handled connection pooling and mem-
ory management more effectively than the old driver.
DataBank solved several issues by deploying a new database driver: loss of
revenue, dissatisfied customers, and overworked IT personnel, to name a few.
Chapter 3 details the many ways that database drivers can affect perfor-
mance and what you can do about it.
To learn how the environment can affect performance, see Figure 1-4 for an
example of how different Java Virtual Machines (JVMs) can cause varied perfor-
mance results for a JDBC-enabled database application. In this example, the
same benchmark application was run three times using the same JDBC driver,
database server, hardware, and operating system. The only variable was the JVM.
The JVMs tested were from different vendors but were the same version and had
comparable configurations. The benchmark measured the throughput and scala-
bility of a database application.
