|
|
Core Routing Table Growth |
The current size and growth rate of the core BGP route table is important to both network operators and researchers, as it shows the impact of network engineering techniques on the limited resources available inside routers in the Internet core. Other studies [Bates, Huston], show the size of the core route table from locations in San Jose, CA USA, and from Australia, and these two measurements have revealed significant differences between the views from these two sites.
The University of Oregon Route Views project [Meyer] maintains a `looking glass' route server that has BGP sessions with many routers spread throughout the world. This makes it an ideal resource for studying the variations in core route table size seen at different locations in the network, and also allows easy comparison of the growth rates seen at those sites.
In the graph below, the number of routes received from each Route Views peer is graphed as a separate line. The color of each line is arbitrary, and is only used to distinguish the number of routes advertised by different peers. Sudden drops in the number of routes from any particular peer are most likely due to problems with the peering session with the Route Views server or the data collection programs, and do not reflect changes in reachability through any of the participant ISPs. The single line extending back past November 1998 is the data collected by Geoff Huston at Telstra in Australia [Huston], and is included for comparison with the data collected from Route Views.
The most interesting features of the graph below are the increasing growth rate for all peers, and the small differences in growth rates among peers that persist over timescales of a year or more. There are roughly three groups of lines, and the majority of the peers belong to the middle group. All of these groups show a similar increase in the rate at which new prefixes are added, but there are small differences in growth rate among the groups causing the separation among the lines to increase with time. More work needs to be done to establish the origins of these effects, but the observed growth rate is clearly larger than the rate at which new IP address space is being allocated. This suggests that the trend is caused by traffic engineering techniques rather than growth in the network.
Hans-Werner Braun maintained an online archive of BGP table snapshots collected from Route Views once per day [Braun] from the fall of 1997 through December of 2000. This archive is the source of our historical data, and is supplemented by current data collected by PCH.