The (un?)reliability of the Internet
I’ve been tinkering with some of the tools I’ve been working on. I have my Traceroute Mesh tool which I wrote ages ago. With some help from AJ from Maxnet, I got a BGP feed, and started doing some more interesting things. I wrote a simple Whois Interface, with syntax colour highlighting, and information live from BGP. It’s kinda useful to see what prefixes are announced by an AS, and what AS announces what prefix, as well as looking up abuse information. I had fun writing this, especially implementing the leaky bucket rate limiting code.
I also wrote a tool that takes a destination prefix/AS, and shows the all the paths I’ve ever seen to that prefix/AS (astr). This can be used when the Internet is “broken” to figure out how the internet actually works when it’s working, or to look at historic data. Yay. However it gave me an idea, looking at almost any prefix it shows multiple paths. So how unstable is the internet?
Well, the answer seems to be “very” unstable. For each prefix I measured the longest time it had between route changes, and plotted this. This shows for instance that 50% of prefixes change more frequently than once every 2 days.
