How We Created a Globally Distributed Monitoring System with PRTG for Less Than US$220/Month

 Originally published on October 08, 2008 by Dirk Paessler
Last updated on March 03, 2022 • 7 minute read

As a testbed and demonstration website for the "Remote Probe" feature of our monitoring software PRTG Network Monitor (and out of technical curiosity) we have set up a network of monitoring stations around the globe based on PRTG. To keep our costs low we have used virtual server offerings from grid hosting or VPS hosting companies for our remote probe installations. Virtual servers are much cheaper than dedicated offerings but provide only limited resources of course. In order to act as a remote probe for PRTG that only sends out a few monitoring requests into the Internet every few seconds a VPS is perfectly enough - in all cases we purchased the smallest Windows VPS offering available and saw decent results. PRTG's remote probes usually require little CPU cycles. We have to keep in mind that - due to bottlenecks caused by other VPSs on the same hosting system that may use a lot of CPU or bandwidth - the measured values will show a lot more jitter than they would on a dedicated system. For a long term perspective these fluctuations do not really matter. And they also do not matter for pure availability monitoring (even if the probe system has little CPU power, it will still find out whether a webserver can be reached or not). As a result we now have a network of 5 monitoring stations around the globe that are connected to one core instance of PRTG Network Monitor.

 

 

We used the following locations and hosting companies:

Location Hosting Company Monthly Cost (US$) Reliability Rating
San Francisco USA GoGrid.com $75 *****
London UK webhosting.co.uk $43 ****
Edinburgh UK FlexiScale.com ($105) *
Cologne DE HostEurope.de $27 *****
Singapore SG usonyx.net $75 ****
Nuremberg DE Our own office datacenter $0  
In my subjective experience the services of GoGrid and HostEurope look like the best offerings. Both didn't have a single downtime in the last 3 months. For the probe in the UK we started with Flexiscale, but it was too expensive and they were offline with their complete grid system for 6 days due to a hardware failure in August (apart from that failure the service was ok). So we are now switching to webhosting.co.uk. For Asia we started with a company in Hongkong which proofed unreliable and we are now switching to usonyx.net in Singapore. Other options that will be available later are: As you can see the monthly total for our setup is around US$ 220. With this little investment (ok, plus a PRTG license) we are now able to create an unlimited number of sensors and we can use all of the features of PRTG 7 for our monitoring (multi user accounts, reporting, maps etc.). You could even create such a network with the freeware edition of PRTG (but it will be limited to a total of 100 sensors).

 

Sample Graph 1: Live Map of Request Times of www.paessler.com

This map from PRTG shows graphs with recent measurements of HTTP requests to the homepage of www.paessler.com (located in Dallas, TX) reported by our probes in Cologne, Nuremberg, San Francisco, Singapore and Edinburgh.

 

Sample Graph 2: Historic Graph

The following report from PRTG shows graphs with the hourly averages for HTTP requests to the homepage of www.paessler.com (located in Dallas, TX) reported by our probes in Cologne, Nuremberg, San Francisco and Edinburgh during the month of September. It is interesting to see that despite the fact that all probes run on virtual systems, in the long run the measured values are quite constant. Regarding uptime measurements all probes have the same result of 100%. Click the image to zoom in!

 

Sample Graph 3: Live Data Tables of Request Times

Another sample is this public map from PRTG that shows the download times for some URLs on our servers from the different locations.