Measuring Application Performance on Windows Server and the Key Perfomance Counters or Indicators

The performance counter data helps determining system bottlenecks and fine tune the application performance. We can utilize the "perfmon utility" on the server. Start > Run > Perfmon and this brings up the perfmon console. Other testing tools which can gather performance counters are Load Runner, VSTS, ...

Steps to measure perfomance counters using out of  box "Perfmon":

1) Download the following "Performance Monitor Wizard" on your server

2) Execute the wizard by double clicking it, give it a name, select terminal server, select all counters, specify path where to save log file and finish the wizard without starting it. So, now we have configured a perfmon template.

3) Open Perfmon, when you want to collect performance counters

4) Start and Stop the configured perfmon (lets say in our case configured template name is "t") as shown below.

Analyze the collected data i.e. performance counters

In the perfmon, do the following as shown in the below image

Click on the data tab and select the counters as shown below:

Now the graph below shows the performance counter selected:

Few Important Key Performance Indicators or Performance Counters to Monitor

Object                                                   Performance                                   Counter Indicates

Processor                                             % Processor Time/_Total                   Processor use level


Memory                                               Available Bytes                                  Amount of memory available


Network                                              Interface Bytes Total/sec                    Amount of network traffic traveling to & from test client
                                                                                                                      (Network contention).


Memory                                              Memory\Pages/sec                             The number of paging operations to disk during the 
                                                                                                                      measuring interval, and this is the primary counter to watch
                                                                                                                      for indication of possible insufficient RAM to meet your
                                                                                                                      server's needs.



The counters mentioned above can help to find out the possibly the bottlenecks and can help in deciding whether to scale up or scale wide. Generally, thumb rule is processor on an average during peak hours should not exceed 60% and page fault should not be more then 50 per seconds.