Developer Documentation Library > Interviewer - Server > Monitoring and tuning system performance > UNICOM Intelligence Load Tool > Guidelines for using UNICOM Intelligence Load Tool
 
Guidelines for using UNICOM Intelligence Load Tool
General guidelines
Installation
Do not install UNICOM Intelligence Load Tool on a server in the UNICOM Intelligence Interviewer cluster or run it from a clustered server. UNICOM Intelligence Load Tool uses significant CPU and memory resources and, ideally, should be run from a standalone server.
Maximum concurrent interviews
Do not run more than 400 interviews simultaneously on a single computer. This number depends on the specifications of the computer running the UNICOM Intelligence Load Tool. If you are running a number of tests at the same time, do not run more than 400 simultaneous interviews altogether across all the tests.
Stagger interview start-up
For a more realistic test, set a delay of at least one second between interviews. This prevents the cluster from returning server busy errors as interview start-up limits are exceeded.
Performance and server load guidelines
Rendering HTML output
Switch Render HTML output off if you are only interested in checking routing under load and not the overall page-to-page performance. This will reduce the Web Tier server load.
Don’t display Info messages
For improved performance, turn off Info logging and just keep Error and Warning selected. Also, with Info logging selected, log entries build up very quickly on the screen, making it difficult to see the actual errors.
Answer pooling
Use Answer Pooling to improve the performance of the UNICOM Intelligence Load Tool and to increase load on the server. Do this only when the loading of the survey is more important than selecting truly random responses.
Respondent guidelines
Respondent navigations
When setting % Stopped, be aware that navigation percentages are evaluated for each page. For example, if an interview has ten questions and you specify a 10% chance of click the Stop button on each page, you would expect that you will hit the Stop button on one of the ten questions, meaning that you will almost always press the Stop button for the interview. Therefore, almost all of your interviews will end up in the Stopped by respondent state.
If you want only a set percentage of all respondents to select the Stop navigation, set % Stopped to that percentage divided by the average number questions in the interview. For example, if the interview has ten questions and you want 10% of respondents to select Stop, set % Stopped to 1% (10/10 questions).
Consistent answers between runs
If you want to rerun a test and have the Auto Answer player select identical sets of answers in both runs, you must set a random interview seed. Selecting this option means that all data collected during the load run should be the same in subsequent runs. This is useful when attempting to reproduce errors.
Simulating browser time-out
To simulate restarts after a time-out from the browser, set the Timeout on the Interview Parameters tab to a value less than the maximum delay between pages set on the Respondent tab.
Hints
Use hints to answer questions with custom validation or if you want to run a specific path through the routing script. When you run the UNICOM Intelligence Load Tool, you can confirm that a hint has been used by looking at the Info log messages in the Logging tab or the MRI*.tmp log files.
Page delays
Use Set delay per question on the Respondent tab to give more realistic wait times for each page in the interview. With this option selected, the wait time is increased when there are more questions in the page to be answered.
See also
UNICOM Intelligence Load Tool