non-rate limited, automatable Looking Glasses?

Mailman List Mirror (Read Only)
1

Hi everyone,

In the next couple of months, I want to compare data plane and control plane measurements on a larger scale. In particular, I'm looking for (publicly accessible) devices that receive BGP feeds and can perform a bunch of automated (paris) traceroutes. I currently do not have an exact probing rate or target set in mind; however, I'm sure that manually entering IP addresses as targets for usual Looking glasses won't cut it. Does anyone know less-restricted (maybe even automatable?) Looking Glasses (or similar devices) or is willing to provide access to one?

BTW: I though about picking Atlas probes from ASes that feed BGP Collector Projects (e.g. RIPE RIS or RouteViews). Unfortunately, the respective probes are often really far apart from the feeding routers; thus, their individual perspectives are likely misaligned :frowning:

Best regards,

Lars

2

Hi Lars,

You should check out https://ring.nlnog.net/ by contributing resources yourself you also get access to a wide array of machines from all across the world you can use to turn traceroutes and pings.

Some wrappers have already been made to run commands against multiple machines at the same time (https://ring.nlnog.net/toolbox/), you’ll have SSH access to run any commands you want and there is an API to find the probes if you want to automate it all.

I encourage anyone and everyone to join. The more networks the better!

Brendan

3

And you can also use ripe atlas as well. If you need credits ask on that list and people offer them up regularly and quickly.

4

Hey Lars,

We’re already doing something similar to this on our collectors, in collaboration with CAIDA. Feel free to contact me off list if you’re interested.

  • David
5

Hi Brendan,

Thanks for the pointer! Unfortunately, not all devices that can perform traceroutes are suitable for my issue; I can only use those that also allow me to inspect their routing table so I can compare the chosen control plane path with the actual data plane path. In addition, the richer a device’s routing table the better—there is not much to measure with only a default route :slight_smile:

Best regards,

Lars

6

Hi Jared,

As I hinted at in the initial mail: The Atlas probes of an AS are often far apart from the BGP-feeding routers of the same AS; thus, it’s unlikely that both devices share the same control plane information which makes it pointless to measure differences.

Best regards,

Lars

7

Just trying to clarify the question. If you observe a BGP route to 1.2.3.0/24 with AS path 1 2 3, you want to do a traceroute to confirm that the packets indeed travel through ASNs 1, 2 and 3?

I would think that traceroute will have to be run directly on the same router that provides the BGP feed.

Regards

Baldur

lør. 18. jul. 2020 23.34 skrev Lars Prehn <lprehn@mpi-inf.mpg.de>:

8

Hi Baldur,

Yes, you are right. While, in general, Looking Glasses would be optimal, those LGs that I know have rules in place that prohibit automated requests and also limit the number of queries one can enter manually.

Best regards,

Lars