Performance Engineering Best Practices and Methodology

[Best Practices] Performance Monitoring - Windows Monitoring, part I

Outstanding Contributor.

[Best Practices] Performance Monitoring - Windows Monitoring, part I

Since a great majority of applications used by IT organizations are Windows based, using Performance Center enables you to use Windows operating system performance counters to trace behavior of your application under

We will discuss preselected collections of monitors that consist of default metrics and default thresholds (where applicable), all of which have been researched using best practice data and expertise from various sources
including HP’s operating system administrators, HP’s Professional Services Organization, technical documentation, and books from industry experts.


Modern Windows platforms, starting with Windows 2000 and later, provide various built-in facilities to gather, display, and reuse performance-related information. These facilities use a variety of sampling techniques to
generate interval performance monitoring data that is extremely useful in diagnosing performance problems. They are designed to be efficient enough that you can run them continuously with minimal impact.



Related performance statistics are organized into objects. For example, measurements related to overall processor usage, such as Interrupts/sec and % User Time, are available in the Processor object.
There might be one or more instances of a performance object, where each instance is named so that it is uniquely identified. For example, on a machine with more than one processor, there is more than one instance of each set of processor measurements. Each processor performance counter is associated with a specific named instance of the Processor object. The instance name is a unique identifier for the set of counters related to that instance, as shown below:



The individual performance statistics that are available for each measurement interval are numeric performance counters. Each performance counter you select is uniquely identified by its path, usually in the following syntax:






The Computer_name portion of the path is optional.
For a simple object such as System or Memory that has only a single object instance associated with it, the use the following syntax:




Each counter has a counter type. Knowing the counter type is useful because it indicates how the performance statistic was derived.
Some of the most important counter types are:
➤ Instantaneous counters. Display a simple numeric value of the most recent measurement
➤ Interval counters. Display an activity rate over time
➤ Elapsed time counters. Gathered on an interval basis and cannot be summarized
➤ Averaging counters. Provide average values derived for a given interval



 This post is part of the Performance Monitoring Best Practices series - you may see all of the posts under PerfMonitoring tag.