Corelation between traffic and prefixes

Tue Apr 27 21:52:59 UTC 1999

Has anyone done a (statistical) corelation and regression exercise
between number of announced prefixes from peers, and traffic sent to
them, and if so would they be prepared to publish the consolidated
data (if not the anonymized data set)? Equally is there a corellation
between advertised address space and traffic sent?

Yes, there is an operational reason for this question. Essentially
I'd like to know (a) whether advertized prefixes have a definable
average "value" as well as "cost" (cost in terms of router resources),
and (b) to what extent total prefixes is in practice a good measure
of network size for peering decisions, or whether one really needs
go do the Netflow. For instance on (a) I have noticed one inverse
corelation - networks which are poor at CIDR tend to cause more
aggro and give less value than the equivalent total address space.

I am interested more in statistical data than opinion, the former
of which I have have a lack of, the latter a surfeit.

I will anonymize and summarize useful private replies to this list unless
a preference otherwise is indicated.

-- 
Alex Bligh
GX Networks (formerly Xara Networks)