Tool Mentor: Generating Datapools for
Realistic Performance Test Data
Purpose
This tool mentor describes how to generate datapools to send realistic values
to the server in performance testing. The datapools are generated in Rational
Robot and are then used by Rational LoadTest during performance testing.
Related Rational Unified Process activities:
Overview
When doing performance testing with LoadTest, the actual data that you send
to the server has a great effect on performance. In the real world, where users
have different names, the server has to retrieve and store each different name
in a back-end database. The database has to do a lot of work to retrieve and
store each different name, which is why Web applications may start to get slow.
For example, in an e-commerce site selling merchandise, the user would fill in
his/her name to order the merchandise. If you run a performance test and every
user has the same name, the back-end database has to retrieve and store only one
name. Because the database can put this single name in a very fast area of
memory ù cache ù the reports you run will tell you that you have a very fast
system. (This is known as a "false positive" test result, since the
test is falsely telling you everything will work.)
To generate a more realistic workload, LoadTest uses datapools. Datapools supply
different names to the server, which will give you a realistic idea of
real-world performance. The datapools are generated in Robot, and then used by
LoadTest when the tests are run.
Procedure
The procedure for generating
datapools is found in the Tutorial section of the Getting Started with
PerformanceStudio manual, which can be found on the documentation CD. The
procedure and detailed information about datapools are also found in Chapter 6
of the Using Rational LoadTest manual, which can be found on the
documentation CD.
Also see the following Guidelines for additional related information:
Copyright
⌐ 1987 - 2000 Rational Software Corporation
|