I am looking to reduce the CPU usage on my PRTG probe. My PRTG hardware is a few years old and it's spiking past 85% utilization at least once a day. I have about 400 sensors on 97 devices.

My first thought is to reduce the amount of probes that it's monitoring, so I'm looking for some direction on disabling which probes will benefit me the most.

One one server, for instance, I have PING, CPU LOAD, DISK FREE, MEMORY UTILIZATION, UPTIME, and RDP. Are all of those things being reported at the same time by a single query? On that particular server, UPTIME and MEMORY UTILIZATION aren't as important as the others, so I'd be fine deleting those probes. However if it's all "reported at the same time" as part of a single query I have my doubts that fine-grain tuning of the probes by removing one or two per server will make any difference in CPU utilization.

Any ideas or thoughts? Any particular probe categories that take up more CPU use than others that I should think about removing first? Thanks!

Article Comments

All sensors have a color code in the lower left corner in the list of selectable sensors.

Recommended limits of a specific color code per probe:

  • 1 bar, green, up to 5 000 sensors
  • 2 bars, light green, up to 2 000 sensors
  • 3 bars, yellow, up to 500 sensors
  • 4 bars, red, up to 200 sensors
  • 5 bars, dark red, up to 50 sensors

Typical, SNMP is less performance demanding than WMI.

CPU is normally not a problem, you will need about 150 Kilobyte of RAM per sensor. The biggest bottleneck are network traffic.

So to get better performance on your probe you could rise the interval for polling the devices and tweak timeout values.

- Jens Tore Fremmegaard, ServerParkering AS, Paessler Certified Monitoring Professional -


Jul, 2014 - Permalink

Thanks for the help. We had many Windows Network Card sensors attached to all of our servers, probably via auto-discovery, and those are all rated red. I will start my investigation by reducing the interval on those, or removing them altogether for less critical systems. Thanks again!


Jul, 2014 - Permalink