Using Amazon EC2 Micro Instances For WAN and Website Monitoring

 Originally published on February 11, 2012 by Dirk Paessler
Last updated on March 03, 2022 • 11 minute read

Recently the Amazon EC2 Cloud announced that their "Micro Instances" now support Windows as operating system. Before this announcement you could only use "Small Instances" for Windows machines on the Amazon Cloud. Now Amazon even offers new customers a Micro Instance free of charge for one year.

For details about the free offer, please see the Amazon Web Services Blog.

We have been using the Amazon EC2 cloud for our own monitoring and for PRTG testing extensively since Windows became available in November of 2008 (see blog article Using Amazon's Elastic Compute Cloud (EC2) to Run a Remote Probe for PRTG Network Monitor). In the early days it was quite complicated to set up an instance. Nowadays Amazon offers a nice AJAX based web interface for instance management. BTW, there are people who say: If you use Amazon's web console for instance management, than you are only using EC2 like an old-school hosting provider. Only if you use their APIs you are actually doing real "cloud computing". That may be the case, but do we care...?

Lower Costs Using Micro Instances

Back to the topic: Amazon has lowered their pricing several times since 2008 and now with Micro Instances it becomes even cheaper to run remote probes or even a whole cluster in the cloud. A Micro Instance now only costs between US$ 21 and US$ 26 per month (depending on the geographical location of the data center).

The differences between small and micro instances are:

 

Micro Instance

Small Instance

Monthly Cost

US$ 21–26

US$ 86–108

System Memory

613 MB

1,700 MB

CPU

Up to 2 EC2 Compute Units
(in bursts)

1 EC2 Compute Unit
(reserved)

I/O Performance

Low

Medium

 

Perfect for Network Monitoring

Amazon themselves describe Micro Instances like this: "Micro instances (t1.micro) provide a small amount of consistent CPU resources and allow you to increase CPU capacity in short bursts when additional cycles are available. They are well suited for lower throughput applications and web sites that require additional compute cycles periodically".

This means that these instances are perfect for a monitoring system! A PRTG probes is idle most of the time until a sensor must be scanned which can require a lot of CPU cycles in a short burst!

By the way: Did you know that Paessler AG is an official AWS Solution Provider? With PRTG you can also monitor your EC2 cloud using our Amazon CloudWatch Sensor.

Put To The Test

The technical specifications of these micro instances are good enough to run a remote probe for PRTG with even several hundred sensors. So we recently switched most of our EC2 based probes for our public demo system from small to micro instances which saves us about 75% of the monthly cost.

PRTG's resource usage is so optimized that you can even run a PRTG cluster with several hundred sensors on these Micro Instances. This means that we can run our globally distributed 5-way cluster test system of PRTG for less than US$ 125 per month now!

When we switched 5 probes and 3 cluster nodes on February 8th at noon, we found the following differences:

  • CPU, memory, and disk performance of the Micro Instances seem to be about 3–5 times better than the performance of our previous Small Instances (see Graph 1). This improvement is better than expected, it looks like our instances not only get more compute-units, they are also running on faster CPUs now, likely due to more recent hardware.
  • CPU, memory and disk performance of the Micro Instances are more volatile than those of Small Instances—which we had to expect from Amazon's description (we don't get a reserved CPU for our instance)
  • Ping time measurements have become more stable on the other side. As expected the average Ping times themselves did not change, but there are less "peaks" in the graphs for Pings (see Graph 2). Average Ping time variation is now in the 10–20 msec range, so this is likely caused by more exact time measuring due to the faster CPU, not by an actually faster or slower network.
  • All other sensors (e.g. http, POP3, SMTP, etc.) do not show any notable differences before and after the switch

How to Monitor Globally

If you are a PRTG user and if you need the monitoring perspective from outside of your own network you should consider setting up remote probes on the EC2 cloud. Micro instances make this option amazingly cheap while delivering enough power for reliable monitoring.

 

Want your own global monitoring cluster? Use the freeware of PRTG (which supports up to 100 sensors) and create a cluster with up to five machines that can monitor your systems or your website from five different locations around the globe for a little more than US$ 100 per month! The locations are:

  • US East (Northern Virginia)
  • US West (Oregon)
  • US West (Northern California)
  • EU (Ireland)
  • Asia Pacific (Singapore)
  • Asia Pacific (Tokyo)
  • South America (Sao Paulo)