RE: ip in the ran

[James Kempf <James.Kempf@sun.com>]

Tim,


>> I'm not exactly sure what the problem is here, but I don't think
>> this has anything to do with whether the RAN uses IP or not.
>
>well, i don't know if its allowed but i suppose i'm fishing for a problem

Sure! Fishing is always useful. Never know what you will catch.

>here.  the dimensioning of an IP-based access network with 10,000 radio
>access points seems to exceed various existing Internet deployments.  the
>largest dial network has about 2000 access points, and there's only a few
>that are this large.  I understand many mobile operators have 10k plus
>access points.  I believe the edge router factor (if i'm not misinformed) is
>on the order of hundreds of routers, which might be easily exceeded.
>

Hmm, I think I understand. Since there seems to be no operational
experience in access networks with this many access points, I
agree there is the potential for a problem.

On the other hand, based on theoretical and design considerations,
I can't see any immediate problem (from the bit I know about
routing) as long as there is no requirement to put all the access
points under the same subnet. It should be possible to partition
the access points into subnets, with a single Access Router per
subnet, and have the core organized in a hierarchical manner
to handle the routing efficiently.

Are you saying that in existing edge networks, all 2000 access points
are deployed under the same subnet (sorry for the clueless question,
I'm not familiar with existing deployments)?

>
>I agree with your points here, but I'd still like to see someone work
>through the dimensioning for a "typical" mobile operator and think about
>routing implications for what no doubt is going to be a large network.
>

Agree! So far, I've not seen any published info on deployments with
this number of access points. It would be great to have an Internet
draft, perhaps for BCP track, that discusses this issue.

		jak