Viewing AR System Log Analyzer output

The initial goals of viewing AR System Log Analyzer output are to:

• Verify that the logs captured the activity you wanted to analyze.
• Verify that the output contains statistics worthy of further investigation.
• Identify any pattern of when individual API calls or SQL statements are slow.
• Identify any pattern of which individual API calls or SQL statements are slow.

Before you begin

Before viewing the AR System Log Analyzer output, ensure that you have performed the following tasks:

• Define the performance issue.
• Capture the issue in appropriate logs.
• Run arpreparelog.exe.
• Run AR System Log Analyzer to generate web (HTML) format output.

To view AR System Log Analyzer output

1. Navigate to the directory that AR System Log Analyzer created.
2. Open index.htm in a browser.
3. Click General Statisticsand review the following information:
   • Verify appropriate elapsed time in the log and note start time, end time, and elapsed time.
      The log should be captured during a time that captures the reported performance issue. The general guideline is to capture at least 30 minutes of activity. If the elapsed time is too large, use arpreparelogs to extract a 30-minute time range subset of the logs during which the issue occurred.
   • Note the user count, which provides a high-level indication of whether other user activity might be involved in the reported performance issue.
   • Note API and SQL exceptions, which indicate API or SQL calls that are not completely captured in the log.
      Letting the logs run for a few minutes after the performance issue is reported increases the likelihood that complete results are recorded in the logs.
4. Under API Aggregates, click Top N.
5. Review the values of Execution Time.
   The general guideline for whether the statistics are relevant to the reported performance issue is:
   • The statistics are relevant if Top Nincludes both:
     • Slow API calls in which Execution Time is greater than 5 seconds
     • Normal API calls in which Execution Time is less than 2 seconds.
   • If all of the Top N operations have Execution Time greater than 5 seconds, the reported performance issue might not be listed in the Top N. You can address this in several possible ways:
     • Investigate and resolve the largest performance issues in the log even if they are not relevant to the reported issue.
     • Run AR System Log Analyzer again, and adjust the value of the -n parameter to capture the top 50 or top 100 in Top N.
     • Extract logs based on a specific user name or time range.
6. Review the values of API and Form to identify frequent long-running API calls.
    The general guideline is 20 percent or more of the entries in Top N that indicate the same combination of API and Form is evidence that a slow operation may be at fault. See Investigating-slow-BMC-Remedy-AR-System-API-calls.
7. Review the values of Start Time to identify a common time for the slow API operations.
    The general guideline is 50 percent of the long-running API calls listed in Top N starting within the same one-minute period is evidence that contention or blocking occurred to cause the performance issue.

Example

An administrator is investigating a performance issue when synchronizing from BMC Atrium Discovery and Dependency Mapping to BMC Atrium CMDB. After capturing the issue in logs, running arpreparelogs, and running AR System Log Analyzer, the administrator opens the output file to view the general statistics. The statistics show the log captured 20 minutes of activity, only 5 users were using the system, and 3 API exceptions. This is the expected behavior.

(Click the image to expand it.)

Then, under API Aggregates, the administrator clicks Top N to see the longest running API operations:

(Click the image to expand it.)

Execution Time values show API calls longer than 18 seconds and as little as 1.2 seconds. This is a large range for Execution Time, investigating why some operations took 18 seconds is worthwhile.

The Top N results also show several similar API operations on the same form, GLEWF on BMC.CORE:BMC_Impact_D.

The values of Start Time of these API calls are within one second of each other, so the administrator decides to look at these operations closer to investigate whether some event on the AR System server or at the database server are in contention, or whether the same API call processes more quickly at other times.

The administrator decides to investigate these slow API operations to determine the cause and whether they are responsible for the reported performance issue.

Where to go from here

Investigating-slow-BMC-Remedy-AR-System-API-calls
Analyzing-thread-statistics