Assessing relevance of performance issues
Before you begin
Before you begin, make sure that you have performed the following tasks to identify performance issues:
- View the results of AR System Log Analyzer.
- Investigate slow API operations.
- Analyze thread statistics.
To identify relevant API calls for further investigation
- Under API Aggregates, click by User.
- Scroll through the list looking for the largest values of AVG TIME for the user encountering the issue.
If the combination of API, Form, and User occurs more than 100 times, and the Avg Time is greater than 3 seconds, improving the performance is generally relevant to the user.
To determine whether slow API calls are related to the reported performance issue
- Under API Aggregates, click Top N and view API calls with Start Time consistent with the issue that the user reported.
- Click the line number to view the log entry.
- Check whether the user column indicates the user reporting the issue.
Repeat the procedure for each of the API calls listed in the Top N with Start Time consistent with the issue that the user reported.
- Under API Aggregates, click Exceptions.
- Click the line number to view the log entry.
- Verify whether the User column indicates the user reporting the issue.
- Under API Aggregates, click Exceptions to return to the list of API exceptions.
- Repeat steps 6 – 8 for each API exception.
Example
An administrator identifies a long-running API call by investigating the Top N API calls.
The API call is a GLEWF on BMC.CORE:BMC_Impact_D, which can take more than 18 seconds to complete.
Under API Aggregates, the administrator clicks By Form.
The following results are displayed:
Reviewing the counts of API calls shows that the one that can take a long time was issued 816 times, which was a significant number during the test. The AVG TIME to complete the API call was a reasonable 0.1567 seconds, but the instances when it did take a long time pushed the SUM TIME to be over 2 minutes, compared to 11 seconds to process a similar number of API calls on BMC.CORE:BMC_Dependency.
The administrator concludes this slow API call is relevant to the overall performance issue captured in the logs.