DX Application Performance Management

CA Tuesday Tip: Configuring Percent of Slow Time (POST) for APM CE Introsco

  • 1.  CA Tuesday Tip: Configuring Percent of Slow Time (POST) for APM CE Introsco

    Broadcom Employee
    Posted Feb 15, 2014 12:17 PM

    CA Tuesday Tip: Configuring Percent of Slow Time (POST) for APM CE Introscope Integration

    This month's tip covers the area of problem resolution triage (aka transaction trace) integration. In a previous tip, I already discussed a knowledgebase article on an approach to use for getting the integration to work.

    Reading through the documentation in the Configuration and Administration Guide, one learns the following:

    1) Go to Setup>Introscope Settings to configure the integration.
    2) Once there, you can configure how long a transaction trace can run (Maximum Transaction Trace Session Duration) and the time threshold that must be exceeded to start a transaction trace (Transaction Trace Time Threshold as a percentage of the slow time defect.) Hereafter, it will be referred to as "percent of slow time or POST"
    3) Once a transaction time on the application server (NOT the web server) exceeds the POST value AND the transaction definition is matched, then the transaction trace is initiated and sent to the EM. 
    4) Setting the value too low may negatively impact an application server.
    5) These settings are global and cannot be applied to individual business services or transactions. So you could get transaction traces on some transactions that are not of interest.

    But other than that, many other questions are unanswered:
    1) What are typical POST values during test?
    2) What are typical POST values during production?
    3) What are the implications of setting too high a value?

    Lets explore each of these in turn:
    1) What are typical POST values during test?

    Typically, there are only a few transactions being monitored by APM so the POST setting can be configured very low such as 5-25%. There is more  concern about learning what contributes to a slow response time than the impact to the application server. Note that an APM Proof of Concept could also use the same settings.

    2) What are typical POST values during production?
    These are ususally set to a low value before a production rollout to see if the integration works. Then for a rollout/optimization, the value is raised higher than the default value of 50%. By doing this, one will only get transaction traces on the slowest 5-10% of transactions. So 50-200% could be possible values depending how many transaction traces you wish to obtain.

    3) What are the implications of setting too high a value?
    If the value is set too high, then a transaction trace may not kick off. Typically by looking at the Traces tab in the Investigator can show this. For example, if the longest transaction trace took a few milliseconds according to the Traces tab. Then a percent of slow time resulting in a threshold of 2.5 seconds will never be exceeded. And if never exceeded, then transaction traces will not appear in a APM CE defect.

    Note, I have also submitted a knowledgebase article on this same topic.

    Questions for Discussion
    1) Would you want to see these values configurable at the Business Service or Business Transaction Level?
    2) What typical POST values do you use for testing and production?
    3) What other functionality would you like to see for CEM-Introscope integration?