[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: Geo pros and cons
On Wed, 2003-04-02 at 20:39, Tony Li wrote:
> Yes, there are regions where geo addressing makes sense. Here
> in the SF bay area, where it's hard not to trip over all of
> the fiber that we have lying around, it would play out
> beautifully. Theoretically.
> However, even in this location, where we have multiple
> interconnect points, we can observe that the local ISPs do
> not choose to connect at all of the interconnects. Some
> are at Fix-W, some at Equinix, some at PAIX, etc.... Even
> in this environment we are not yet at the point where geo
> addressing would not require us to force particular links to
> be installed.
This is my question. do you have some data about the real topology that
indicates that even in areas where the interconnection is dense there
are still unconnected peers?
I agree with you in this point and I am certainly not suggesting forcing
the existence of links, just some sort of opportunistic usage of geo
aggregation. In areas where the interconnection exists, evaluate the
possibility of using geo addressing. If within the geo area the
interconnection is partitioned, just don't use it
> I dunno about you, but I don't like pushing string.
> | -----Original Message-----
> | From: marcelo bagnulo [mailto:firstname.lastname@example.org]
> | Sent: Wednesday, April 02, 2003 7:14 AM
> | To: Tony Li
> | Cc: Iljitsch van Beijnum; email@example.com
> | Subject: RE: Geo pros and cons
> | Hi Tony,
> | [...]
> | > As we concluded many years ago: for addressing to scale, it has
> | > to match the topology.
> | Fully agree.
> | I would say that the question is if topology matches
> | geography or not.
> | I have heard many claims that the Internet is becoming a dense
> | interconnection mesh. If this is true, it should be easy
> | to determine
> | geographical areas that are interconnected, so geo
> | aggregation works
> | without the need of more specific routes.
> | I think that this is certainly not the case for all geo
> | locations, but i
> | would say that it is probably the case for some locations
> | as big cities
> | where a lot of users reside. Wouldn't geo addressing be
> | suitable for
> | this situations?
> | I am not aware of any analysis of how dense is the
> | interconnection in
> | geo areas, but i would say that this input is important
> | when considering
> | geo addressing. Does anyone have information about this?
> | Thanks, marcelo
> | > If addressing does not match the topology,
> | > then additional information in the form of longer
> | prefixes must be
> | > advertised into the routing subsystem. Ergo, if one
> | chooses geographic
> | > address, one must force only geographically based links.
> | Anything
> | > else destroys the aggregatability of the address
> | assignment. Since
> | > we, as IETF members, cannot decree where folks will connect, geo
> | > addressing is a nice theorectical goal which is unimplementable.
> | >
> | > Regards,
> | > Tony
> | --
> | marcelo bagnulo <firstname.lastname@example.org>
> | uc3m
marcelo bagnulo <email@example.com>