Transferred from: http://www.cnblogs.com/Carrie_Liang/archive/2008/11/10/1330997.html
The previous article continues, and then the previous one. The previous article talked about how to use assertion to make a preliminary screening of the test results. So, when we get the test results, how should we look at them? And how did they get here?
First, the use of listener
People who have used loadrunner should know that LoadRunner will provide us with a lot of icons and curves. But in JMeter, we can only find a few poor listener to make it easier for us to see the test results. However, for beginners, some simple results analysis tools can make it easier to understand how the performance test results are analyzed. So, don't underestimate these simple listener.
A.aggregate Report Aggregation reports
We can see that through this report we can get some of the most interesting results in the usual sense of performance testing.
Samples-How many transaction have been completed in this scene
Average--Average response time
Median--median of response time above statistical significance
90% Line--90% of all transaction in the response time of transaction is less than XX
Min--Minimum response time
Max--Maximum response time
PS: The units of the above time are MS
Error--Errors rate
Troughput--throughput, unit: transaction/sec
Kb/sec-The throughput measured by traffic
B.view Results tree view results in a TreeView list
With this listener, we can see a very detailed result of each transaction that it returns, where red refers to the transaction of the error, and the green is passed.
If you have a lot of transaction done in your test scenario, it is recommended that you only log the error transaction in this listener. To do this, you just need to log/display: in the Errors hook on it.
Ii. analysis of the. jtl file
In the performance testing process, we often need to save the test results in a file, so that both the test results can be saved, but also for future performance test reports to provide more material.
JMeter, the results are stored in the. jtl file. this. jtl file can be written in a variety of formats, and generally we record it in CSV file format, because the CSV file format looks convenient, and more importantly, it can provide a lot of convenience for two of times.
The two analysis I'm talking about here means that in addition to using listener, we can also analyze the. jtl file again.
A. Setting the JTL file format
The jmeter we download from the official website of JMeter can be used directly after decompression. However, the content of the JTL file generated by using the default configuration does not meet our needs. So we have to make the necessary settings. In version 2.2, if you want to modify the JTL setting you have to set it to the Jmeter.properties file, but in version 2.3 we just need to set it on the interface. You only need to select a listener and click the Configure button on the page. At this point, a set-up interface will pop out, it is recommended to check the following: Save Field name,save assertion Failure Message.
Items in the B.jtl file
After the above settings, the saved JTL file will have the following items:
Timestamp,elapsed,label,responsecode,responsemessage,threadname,datatype,success,failuremessage,bytes,latency
The absolute time of the request, the response time, the requested label, the return code, the return message, the thread to which the request belongs, the data type, whether the success, the failure information, the byte, the response time
Among the aggregated reports, throughput = number of transaction completed/time required to complete these transaction, average response time = sum of all response times/number of transaction completed, failure rate = number of failures/transaction
Warm tip: In jmeter2.2 and 2.3, there is a problem is that when we reopen JMeter, using a listener to view the JTL file, JMeter will be an error. So when you've finished testing a scene using the command line, you get just a bunch of raw data stored in the JTL file. So knowing the source of the aggregation report is a handy way to get rid of the test tools and analyze the results.
In general, for the results of jmeter analysis, mainly on the JTL files in the original data collation, I use a few small scripts for the relevant analysis, do not know what you intend to do?
Anyway, you can always find a way to analyze your own data.
JMeter Response Results Analysis II