Ping server survey, September 2006

About a month and a half ago I mentioned that I was going to start monitoring ‘ping relay servers’ to see what sort of performance and availability they get. There were only three that I could monitor at the time: Autopinger, Blogflux, and Ping-o-matic.

Methodology

To test the ping servers I’ve been sending each of them an XML-RPC ping every ten minutes, delivered from a random test server (from a pool of about 10) connected to the internet via one of three or four Tier 1 carriers. If ever a ping test should fail, I send another ping within a few seconds from a different monitoring server that is connected to the internet via a different Tier 1 carrier. This way I can rule out any hardware or connectivity-related issues at my end, and get more accurate results from the survey. (It’s handy working for a company with these kind of resources )

A good test results in an XML message saying something like “Thanks for the ping“, has a server 200 OK status, and responds within 10 seconds. So for example if I don’t get a response within 10 seconds, then that is considered to be a bad test.

I won’t go into the testing methodology too much now as I have to be up in six hours (and I’m soooo tired!), but if you have questions then leave a message blow or drop me an email.

Survey Results

Availability

^*For example, couldn’t connect to web server, web server failed to respond in time, web server returned an invalid response etc.
^**For example a server 500 error or a DNS outage.

So as you can see, Autopinger had the highest availability with an impressive 99.23% uptime, followed by Ping-o-matic, and them Blogflux. The main problem that Blogflux suffered was response times (see below), as the test would frequently go over the ten second threshold.

Other observations

While uptime is probably the most important thing I can report on, there are a few other things I’ve notice so far, most notably the response times from these services:

• Autopinger was quite variable, with a typical reponse time ranging from 150 to 350 milliseconds.
• Blogflux typically took just over three seconds to respond.
• Ping-o-matic had a typical response time of 150 to 200 milliseconds.

This response time is important because it can affect people using blog software that pings these servers. If they use Ping-o-matic or Autopinger they probably won’t notice any delay while their blog software sends the ping. Users of Blogflux however will have to wait some time while the ping is dispatched and handled.

That’s all for September for now. If I get some time (and if anyone is interested) I may look at some of the other results captured, but as always: don’t hold your breath!

Tags: xml-rpc, ping, survey, autopinger, blogflux, pingomatic