Peering matrix information at IXPs

Daniele Arena daniele.arena at
Wed Jun 13 16:47:30 UTC 2007

Hi Ricardo,

> >I was wondering if there is any web page with pointers to IXP's peering
> >matrices (such as

I don't think that there is a webpage ( that has
links to peering matrices. There are webpages with pointers to IXPs,
and IXPs may or may not have peering matrices.

> Also, I wanted to know how usually these matrices are built - if they are
> produced from IRR data, manually , or from traffic measurements?

Depending on the IXP, any of the above.

IRR data usually only tells you if two entities peer, it does not give
details about where they peer. If they have peerings at two IXPs, you
generally won't be able to know where packets pass by reading IRR data
(the peering point could be specified in RPSL, but people usually do
not bother to go to that level of detail when updating the IRR

Traffic matrices (at least up to now and AFAIK) have only been built
by traffic measurements by IXPs whose switches all support sFlow. Not
all switches do, so not many IXPs have that feature.

Several IXPs have traffic matrices built on data entered manually.
That is usually not too reliable as people forget to update the
peering data after having updated their peerings on the routers.

On 6/13/07, Roland Perry <lists at> wrote:
> Is a mine of useful information. I'm sure someone will know how that's
> generated (sorry, I don't).

This is actually a totally different peering matrix, it does not give
any info on who peers with whom. It only tells you the amount of AS
overlap between IXPs. The idea behind it is that an ISP might want to
choose at which IXP to be present, depending on what IXPs they already
are members of and looking at how many more ASes they could reach.

The matrix is built on manually entered data (every IXP puts in the
list of the ASes present at their site).

Hope that helps!



More information about the NANOG mailing list