APM Metrics

CA APM displays application performance data that is collected from remote and local systems as metrics.
apmdevops102
CA APM displays application performance data that is collected from remote and local systems as metrics.
How APM Provides Metrics
CA APM monitors the application performance of individual methods as various application components execute them.
  1. Probes that are inserted into application component bytecode report data to the agents.
  2. The agents report the data to the Introscope Enterprise Manager. Other subsystems, like JMX and PMI, also report data which agents collect.
  3. The Enterprise Manager compiles this data into metrics -- application performance as measured at many points in the application subsystems. 
    Note:
     The Enterprise Manager records performance time for system events in a performance log file, <
    EM_Home
    >/logs/perflog.txt. As an alternative to the metrics displayed in the Investigator, the perflog.txt file can contain useful information. For more information about perflog.txt, see the .
  4. Metrics display in the user interface:
    Team Center – Displays BlamePoint metrics in the details panel of the map view as sparklines.
     Workstation -- Provides the Investigator, console, and APM Status Console for viewing application health and data.
    WebView -- Presents the customizable dashboards and the Workstation tree views in a browser interface. These capabilities let you view critical information anytime and anywhere.
  5. You can export the metrics to an external database.
Common Terms
To understand metrics, learn how CA APM uses some common terms.
More terms are available in the CA APM Glossary
.
backend
A backend is an external system, such as a database, a mail server, a transaction processing system (such as CICS or Tuxedo), or a messaging system (such as WebSphere MQ).
concurrency and concurrent invocations
Concurrent methods are methods that started during an interval without finishing during the same interval. Because you want methods to complete quickly, an unusually high value for concurrent invocations is undesirable.
errors
Errors that the application or system being monitored generates.
events
An Introscope event is any action for which agents capture data in addition to metrics. Example events include transaction traces, errors, and stalls. Introscope records events in specific situations, which include:
  • Transaction traces
  • Stalls
  • Errors - captures throwing/catching exceptions, and traces all locations where exceptions are thrown and caught.
    Note:
     Turn off exception catching in production because it can cause significant performance degradation.
frontend
A frontend is the component of an application that first handles an incoming request. The component can be a Servlet, a JSP, a management DB, an EJB, or some other component.
harvest
Harvest is the process in which Introscope gathers data from Collectors.
interval
An interval is a user-defined time slice that is used to define and average metrics. In Introscope, this period is usually 7.5 seconds. Some of the monitored systems capture data at a different interval.
response
Response always refers to method execution. Response is measured as:
  • count--the number of transactions finished during that interval.
  • time--the time it took to execute a method, in milliseconds.
Responses Per Interval is the standard Introscope throughput metric.
response time
Response time is the period to execute a method that is measured as:
  • average response time (ms) -- The average time, in milliseconds, it took to execute the method during the interval.
  • response time, minimum, and maximum -- The lowest and highest response times during the interval.
rate
Rate is the number of method executions per second or time interval.
stall
A stall is an instance where the invocation time for a method has exceeded a threshold that an administrator defined.
Types of Metrics
Types of metrics include:
Count Metrics
Count is an integer. For example, count can represent:
  • The number of data points which were averaged to compute a metric.
  • The number of events since a certain point in time
  • The number of threads in use
Examples of count metrics are errors and stall count.
Heuristic Metrics
Heuristic metrics evaluate and report status. Heuristic metrics are integers. The integers are symbols of status and do not measure anything. For example, a dashboard alert can be based on a heuristic metric with these values:
0 = green = normal
1 = yellow = caution
2 = red = danger
These values are only examples. You can configure your system with different values.
Percentage Metrics
Percentages
 are used to measure resource use against the maximum available resources. Examples are:
  • CPU utilization
  • Percentage of time that is spent in Garbage Collections during the last 15 minutes
String Data
In addition to measurements and status, Introscope collects information that identifies monitored applications and systems. Examples of this type of data are system component names such as the name of a database, JVM versions, or IP address.