mat-wg

Download

mat-wg@ripe.net

May 2013

10 participants
11 discussions

Re: [mat-wg] Full BGP full view
by Pietro Marchetta 02 May '13

02 May '13

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

2 1