Route object stats

[SichangHe]

I think we should try to provide more info on the `route' objects:

• [x] how many route objects with different origins for the same prefix?
• [x] how many route objects defined by not all the same maintainers for the same prefix?
• [x] how many route objects defined by not some the same maintainers for the same prefix?
• [x] Collect all routes and put on pugna, @SichangHe They are at pugna:~/internet_route_verification/route_verification/stat_route_objects/aggregated_route_objects.json.gz, @cunha.
• [ ] a bit more complicated, but another anaysis we can do is to compute the set of prefixes with stable origins (see DISCO: Sidestepping RPKI’s Deployment Barriers), and then check how well the route objects match these prefixes.

[SichangHe]

With 93d2c28:

[2024-03-24T03:49:42Z DEBUG stat_route_objects] Scanned `["../../data/irrs/backup", "../../data/irrs/priority/"]`.
[2024-03-24T03:49:44Z DEBUG stat_route_objects] Aggregated 2817344 routes.
[2024-03-24T03:49:44Z DEBUG stat_route_objects] 697269 routes defined multiple times.
[2024-03-24T03:49:44Z DEBUG stat_route_objects] 404901 routes with different origins.
[2024-03-24T03:49:44Z DEBUG stat_route_objects] 469003 routes defined by multiple maintainers.

route_objects_defined_multiple_times.json.gz

[SichangHe]

@cunha, I am counting "different maintainers" by any difference in the aggregated list of mnt-by, since mnt-by can be defined multiple times and comma-separated.

[cunha]

This is the "worst-case" scenario (as in we would have an upper bound on the number of different maintainers), it would also be nice to get the "best-case" scenario (lower bound) by considering any overlap in the list of mnt-by as the same maintainer.

On Sun, Mar 24, 2024 at 12:52 AM Steven Hé @.***> wrote:

@cunha https://github.com/cunha, I am counting "different maintainers" by any difference in the aggregated list of mnt-by, since mnt-by can be defined multiple times and comma-separated.

— Reply to this email directly, view it on GitHub https://github.com/SichangHe/internet_route_verification/issues/138#issuecomment-2016683320, or unsubscribe https://github.com/notifications/unsubscribe-auth/AACPO565QR666REUDF7IXTDYZZEWDAVCNFSM6AAAAABCLRTVSWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAMJWGY4DGMZSGA . You are receiving this because you were mentioned.Message ID: @.***>

[SichangHe]

Updated: With 984ce0650c02f2e84bd41c24e89ec013bef190d2:

[2024-04-09T10:40:47Z DEBUG stat_route_objects] Scanned `["../../data/irrs/backup", "../../data/irrs/priority/"]`.
[2024-04-09T10:40:48Z DEBUG stat_route_objects] Aggregated 2817344 routes.
[2024-04-09T10:40:48Z DEBUG stat_route_objects] 697269 routes defined multiple times.
[2024-04-09T10:40:48Z DEBUG stat_route_objects] 404901 routes with different origins.
[2024-04-09T10:40:48Z DEBUG stat_route_objects] 469003 routes defined by different (not entirely the same) maintainers.
[2024-04-09T10:40:49Z DEBUG stat_route_objects] 465072 routes defined without a common maintainer.

It seems that the difference between #routes defined by non-overlapping maintainers and #routes defined by not entirely-overlapping maintainers is little (around 4000 out of 470k), @cunha.

[cunha]

I find these stats a bit confusing. Maybe there's a more direct way of expressing them.

Could we say that we have:

• 2817344 routes
• 697269 prefixes
• 404901 origins
• 469003 overlapping maintainer sets (for route objects)
• 465072 non-overlapping maintainer sets (for route objects)

One of the issues I am having right now is this: if "469003 routes defined by different (not entirely the same) maintainers", then what are the other routes defined by?

[SichangHe]

Routes are at pugna:~/internet_route_verification/route_verification/stat_route_objects/aggregated_route_objects.json.gz, @cunha.

[SichangHe]

3,904,352 (2,120,075 + 1,784,277) route objects for 2,120,075 prefixes defined only once and 697,269 prefixes defined multiple times.