CAIDA adds geo-tags to its Internet AS dataset
310,000 network links: that's how big your Interwebs grew. Mapping them is very useful
We know what you're thinking, “when did a label on a diagram rate a news story?” but bear with us: the 'net-boffins at CAIDA have added labels to the vast dataset that maps AS-AS links worldwide.
Autonomous systems (AS) are collections of IP addresses administered by a single entity. There are lots of ASes in the world – more than 55,000 according to the CIDR report and the biggest is naturally in China with an IP address span of 120,775,936 – so merely gathering data on their connectivity is no small feat.
To do the geo-labelling (discussed here), the Center for Applied Internet Data Analysis (CAIDA) was careful to maintain “the simple node and link structure of the AS graph, and allows us to capture some of the geographic complexity in the topology”.
Network architects need to make decisions about how they want traffic routed – for example, to choose between the lowest-cost link and the shortest path.
The example CAIDA gives is the route between two universities (The University California, San Diego and the University of Washington): without geographic path labelling, a path via Cogent (AS 174) or via NTT (AS 2914) look the same.
However, if one AS is in Las Vegas and the other is in Los Angeles, the LA route will have a little less latency.
The other reason the geo-label is useful is it makes it easier to identify ASes with redundancy.
CIADA says its geographical AS data was compiled from BGP communities, Looking Glass servers, peering info, and the group's Macroscopic Internet Topology Data Kit.
Oh, and in case you're wondering: there are more than 310,000 AS-AS links in the data set. ®