claes.JPG (2225 bytes)

everGrow.JPG (3626 bytes)

DIMES/CAIDA comparison

The tools has many modalities of usage. We show here how it can be used to compare DIMES and CAIDA data for the month of November 2006.

In makeing the comparison, we use DIMES' AS edge data for the month of November, while for CAIDA's skitter we merged all the daily AS edge files that were released for that month. We used our tool to check if the AS nodes, in either data set, do exist in any of RIR's publically avaliable databases, or have undergone any changes during the period in question. This analysis then results in a reduced set, which we then use for doing futher analysis. The table below shows the number of nodes.
Table 1: AS data from DIMES and CAIDA for the month of November 2006.
  CAIDA DIMES
  # of nodes # of edges # of nodes # of edges
Total 8095 22467 17075 51224
Resolved 6848 17940 15580 45332
DE 730 1279 1368 2598
ES 188 273 997 1184
FR 275 386 475 867
IT 220  328 421 711

The figure below shows the the within country AS connectivity of three countries. It clear from figure that DIMES reveals a much deeper connectivity Structure. Those graphs can be interactively probed using the tool and one can reveal the AS information of any node on the graph.


Figure 1: Within country AS connectivity for Germany, France, and Italy
Within country AS connectivity

K-shell analysis helps uncover information about the role of nodes in a graph in a manner far more revealing than in terms of their degree distribution. Using k-shell analysis [3], the tool can be used for visualizing2 the k-shell graph for the DIMES data as in Figure 2. As in the previous figure, one can interactively probe the AS nodes and extract their detailed ownership and location information from the RIRs’s “whois” databases. The tool can also be used to compare between different k-shell distribution. In Figure 3 we show how DIMES’s k-shell distribution is altered by pruning it to contain on the valid edges. One surmise from the figure that the change in k-shell distribution was relatively mild, with the exception of the innermost k-core.

Figure 2: k-shell layout of DIME’s November 2006 AS data
k-shelllayout

Figure 3: Changes in k-shell distributions
k-shell distributions