For one of the projects a .net application had 10 servers, in addition to monitoring individual usage of resources, we were monitoring cumulative usage of memory and cpu for all 10 servers..
When we increased load by 20% we were expecting an increase in usage of resources…cumulative graphs helped here to determine whether there was an overall increase in cpu and memory usage when all 10 servers are seen as one unit..
We observed cpu and memory usage did not increase despite having around 50% of free resources. There was some degradation in performance observed as well..
This raised a question whether any resource usage limit was placed on application server..on checking app pools it was discovered that indeed cpu usage limit was placed on app pools…
Cumulative graphs helped us find this configuration bottleneck in .net application.
We reran the test after removing cpu usage limit from app pools. We could now observe increase in cumulative cpu and memory usage as per increase in load.