|
The next step after recording a business case is to configure multiple business
cases into a "load profile". Although optional, using load profiles greatly increases the accuracy of
your test, since it better simulates what your web server will be seeing in the real world. In terms of back-end
software, running several tests simultaneously is great for looking for locking problems either in threaded application
code or in database access.
A load profile consists of one or more business cases, each with unique settings
for such things as modem speed, repeat delay, etc.
Create a Load Profile
If no load profiles exist, click on a business case, and then click on the "left arrow" button. A new
load profile will be created automatically for you. When you need to create subsequent load profiles use the Edit->Create->Load
Profile menu item, and a new auto numbered load profile be created. To edit the name of a load profile, first
select the load profile, and then execute Edit->;Edit Load Profile Name.
Assign Business Cases
Load profiles do nothing by themselves; you have to add business cases to them which will be played back during
a performance test.
To add a business case to a load profile, use the following steps:
- Select a load profile by clicking on one in the load profile list.
- Select a business case by clicking on one in the business case list.
- Click the left facing arrow.
The business case you selected is now in the load profile, and you can edit the
business case's settings:
| Host |
All of the URLs contained in the recorded business case are retargeted at runtime
at a new host. This enables you to run the recorded business cases at different hosts without having to re-record
them. If you have a business case that contains multiple hosts, however, you should use the HTTP editing feature
to change hosts.
If you also need to set the port number, use the syntax host:port. Example: 192.168.1.103:81. Use an IP address
rather than a host name for the best performance. |
| Repeat Delay |
A virtual user is assigned to play back a single business case repeatedly. The repeat
delay setting is the number of milliseconds to delay between repeats. |
| Browser |
The browser field of the recorded HTTP request headers are changed at playback time
to contain the string typed into this field. |
| Percentage |
The percentage of total virtual users that will be assigned to this business case.
If you have two business cases set to 50% each, and the performance test starts out with 10 virtual users, 5 users
will be assigned to each of the business cases. As the number of virtual users increases, they will be assigned
to the business cases according to the percentages, keeping the correct ratio. |
| Pacing |
There are two states for this option, "consecutive" and "recorded".
When "consecutive" is chosen the web pages in business cases are played back in such a way as to create
a steady byte rate as determined by the modem speed. Repeats are then subject to a repeat delay.
When "recorded" is chosen, the modem speed has no affect, and the web pages are played back at the speed
they were recorded. For example, if the user paused for 10 minutes while recorded the original business case, the
virtual users would pause at the same place for 10 minutes before playing back the next web page. |
| Random Start |
When this option is selected virtual users do not start playing back at the same
time, but are distributed over a period of a minute. This option makes better use of the available CPU time. Take
a case where all of the virtual users were playing back a business case that took 30 seconds to execute, and they
had a 30 second repeat delay. The CPU would be idle for 30 seconds, and then would loaded down for 30 seconds.
By distributing playback of the business cases over the available time period you can simulate many more virtual
users.
The default state for this option is on. When it is turned off, all of the virtual users start simultaneously. |
| Modem Speed |
This feature
allows Web Performance TrainerTM to simulate the browser connecting
to the web server over different network speeds, from a 28.8kbps
modem to a 100mbps LAN. The parameters are in bits per second,
taking into account the two stop bits for 28.8kbps through
56kbps settings.
What this setting does is throttle the data the simulated user can read or write data to simulate data transfer
through a modem or through a LAN. This affects the back end software because connections are kept open for a much
greater length of time. For example, if your application generates a 40K graph, the browser might spend a fraction
of a section to read the graph when run on the same machine as the server, but could take up to 13 seconds when
the browser is connecting over a modem. Having the socket open for 13 seconds instead of a fraction of a section
puts a greater burden on the server in terms of how many sockets are kept open at one time, and how much memory
is needed to service all of the HTTP requests. |
Multiple Load Profiles
As you create and use more than one load profile it can sometimes be easy to confuse which one you're editing.
Remember that the selected load profile and business case is the one you're editing. If no business case is selected,
you aren't editing one.
|
