[mat-wg] Full BGP full view

Dear,

while a complete and exhaustive view of the BGP global routing would be a
milestone for the whole networking research, all the available datasets
provide just a partial view, since,  as already pointed out by Sebastian
Castro, the current datasets are typically obtanied through a set of
vantage points with a limited visibility.

You might find a recent and well-done analysis on the incompleteness of the
BGP view  as well as the root causes in this paper:

Enrico Gregori, Alessandro Improta, Luciano Lenzini, Lorenzo Rossi, and
Luca Sani.  On the incompleteness of the AS-level graph: a novel
methodology for BGP route collector placement. IMC 2012

At the same time, there several severe sources of inaccuracy to take into
account when you use Traceroute to infer the AS level
topology/transitions/neighborhood.

A good starting reference in this case could be the following paper:

Yu Zhang, Ricardo V. Oliveira, Hongli Zhang, Lixia Zhang.
Quantifying the Pitfalls of Traceroute in AS Connectivity Inference.
PAM 2010

Furthermore, we have recently demonstrated how Third-party addresses cause
the inference of a significant percentage of false AS-level links and bogus
AS-level loops when Traceroute is used to investigate the AS
level connectivity. While this specific limitation has been largely
ignored, you may find many more details in this paper:

Pietro Marchetta, Walter de Donato, Antonio Pescapé.
Detecting Third-party Addresses in Traceroute Traces with IP Timestamp
Option.
PAM 2013.

I hope this helps.
Best regards.
Pietro Marchetta


On 01/05/13 00:43, Pavel V. Veselovskiy wrote:

Hello,


Hi,



I'm validating a new technique for finding the route between autonomous

systems with the least number

of transitions. This method is came from the wireless sensor networks.


To compare the efficiency of this method, and BGP algorithms, we need a

full BGP routing table that reflects the configuration of all the

neighbors for all AS on the Internet. At first I took the BGP Full View

from O-IX site http://www.routeviews.org/

And that's specifically from here

http://archive.routeviews.org/oix-route-views/


Already in the first few tests, it was found out that these tables are

incomplete, since a number of traceroutes, we have given from the

servers is not suitable neighbors and a method results mismatched with the
actual path, or the path is

longer than the traceroute outputs. Besides knowing the total number of

announced AS and counting the number of unique AS numbers for these log

files (http://archive.routeviews.org/oix-route-views/), we found that

these logs contains references to only about 60% of the number All

advertised by AS.



I've came across the same issue while doing some research for CAIDA. You
have to bear in mind the locations where the BGP dumps are taken are
"vantage point", so you will get a view of the network from that point.
If you collect enough vantage point, you will reach a state when you can
"see" mostly everything.

What I did at that time was combine various sources from RouteViews and
RIPE RIS (http://www.ripe.net/data-tools/stats/ris/ris-raw-data) until
reaching an acceptable level of coverage.


Recommend, please, where we can find a really complete picture of all

the neighboring AS?



Cheers,


--

Kind Regards,

Pavel Veselovskiy,

Samara State Aerospace University




-- 
Sebastian Castro
DNS Specialist
.nz Registry Services (New Zealand Domain Name Registry Limited)
desk: +64 4 495 2337
mobile: +64 21 400535


-- 
Pietro Marchetta, PhD student
Dipartimento di Ingegneria Elettrica e delle Tecnologie dell'Informazione
University of Napoli ''Federico II''
Via Claudio, 21 - 80125, Napoli (Italy)
WebSite: http://wpage.unina.it/pietro.marchetta
Phone: +390817683821 - Fax: +390817683816
-------------- next part --------------
An HTML attachment was scrubbed...
URL: </ripe/mail/archives/mat-wg/attachments/20130502/3f16a689/attachment.html>