Performance Testing - Software Testing: Testing Across the Entire Software Development Life Cycle - page 136

The purpose of doing performance testing on components is to get an early idea
of whether the sum of the individual component response times can be expected to
come anywhere close to the performance response time maximum we have recorded
in our peak workload plans. If, for example, the worst database search response time
for our invoice item display is 6 seconds and the worst screen rendering for that item's
description is 5 seconds, then you can alert the developer that the current version of the
Invoice Item Detail Display code has an 11 seconds total response time. If the business
performance requirement for this display is 7 seconds, you have just given the developer
an early warning that his or her code as is will not meet the performance requirements.
9.5 ROUNDTRIPPERFORMANCE
Other meanings of the term “performance testing” tend to encompass more and more
of the application activities to be included in response time measurement. At the far
end of the performance testing defi nition spectrum is the measurement of the appli-
cation's response time from when the “enter” key or “submit” button is pressed until
the results have been fully rendered on the screen or until the report has been printed
out or when the product is delivered to your door. Although there are several terms
in the literature for this kind of performance testing, we will use the term “round trip
performance testing.” When we use this term, we will explicitly mean the time that
it takes for one transaction or activity to be initiated by the user pressing Enter or
Submit until the complete result such as a screen, report, or fi le is returned to the user.
This overall timing will include all intervening processing done on the user's com-
puter, all necessary communications with other support computers via networks, and
all secondary processing done on other support computers for a given transaction.
To demonstrate what round trip performance testing might look like, consider
putting timing points at various strategic points in the test environment platform
of the application such as on the client computer, on the network, and on the server
computer. Then, consider running just one purchase step transaction in the empty
test environment. For the purposes of discussion, say the completed round trip took
a total of 5.8 seconds and was comprised of the following individual timings:
Purchase step transaction round trip performance instrumentation
0.5 seconds client—purchase screen data input record generation
0.2 seconds network—transmission of new purchase record to server
2.4 seconds server—new database purchase record insert into database
1.3 seconds server—generate warehouse order record
0.7 seconds server—generate purchase order confi rmation record
0.1 seconds network - transmission of confi rmation record to client
0.6 seconds client - display confi rmation record / successful completion
=====
5.8 seconds total round trip performance response time in an empty test system