[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: geo short vs long term? [Re: Geo pros and cons]

On donderdag, apr 3, 2003, at 12:51 Europe/Amsterdam, Pekka Savola wrote:

I would not classify geo as a short-term solution.

Short to intermediate term.
Sure, it can be implemented rather quickly, no doubt about that. For the
first years "geo" is probably just a shorthand for "advertise full /48's
from under specific geo-prefix".
Agree.


And in 3-5 years when we try to start
aggregating, we might run into troubles.  And what do we do if we can't
solve those problems sufficiently?
We have to make sure we know in advance that we can handle all reasonable eventualities.

But can it be unimplemented if that's seen fit?  Doesn't seem likely.
The idea is that we create an upgrade path. Geographically aggregatable PI addresses could be used as identifiers in an identifier/locator separation solution. That way, we can get rid of the geo /48s in the routing table without the need for renumbering.

From your previous message:

Similar seems to apply to geo-like approaches depending on geo-aggregates.
Who (at the originating region) is creating the aggregate? And more
importantly, *why* should the sites or ISP's in the region be encouraged
to *not* advertise their more specific geo prefix *anyway* (assuming the
geo routing infrastructure might require that under some conditions)?
ISPs are _supposed_ to announce all geo /48s for their customers everywhere. Aggregation happens inside each individual ISP network by creating private aggregates and selectively filtering geo /48s in certain routers. The only way for peer networks to frustrate aggration is to not interconnect or not announce geo /48s in the expected interconnect location.

It seems likely that the largest networks will gravitate towards a common level of aggregation that conforms to the level of interconnection between those networks. Successfully aggregating more aggressively than this will be hard, and not announcing geo /48s in the closest place you interconnect makes no sense.

When this has happened, smaller networks will be forced to play along and interconnect there where the large networks expect them to because large networks often can't be bothered to peer anyway, and breaking aggregation isn't likely to be a plus.

For smaller networks aggregating is simple, as they can always dump traffic for which they have aggregated away the /48s in one of their transit's laps.