Если вы используете управляемое размещение, эта статья к вам не относится.
System performance can be defined as service time plus latency. When interfaces begin to queue requests, latency occurs until the bottleneck no longer exists. Blackboard recommends measuring the state of the interfaces on a regular basis using an instrumentation tool set such as Server-Status, or predefining measurements using a tool such as Apache JMeter.
Taking crude measurements at an operating system level using tools such as NetStat can provide similar details. In addition, APM tool like New Relic, AppDynamics, and dynaTrace can provide historical metrics for the web server interfaces.