> As I write above, the experiment does not portray the
full extent of
> routing and OSPF. The scalability problems (routes
must be flooded to all
> routers) mentioned in the report don't need to exist:
the report is
> clearly based on a simple topology, but other
topologies such as using
> Stub areas eliminate that specific scalability concern.
>
> FWIW, the reason why I think re-using a routing
protocol may be a good
> choice is that it allows a lot of deployment
flexibility. In scenarios
> where you have a smaller number of nodes or access
routers, flooding (or
> some other distribution) of the routing updates (which
removes the LMA
> single point of failure and optimizes the traffic flow)
requires little or
> no additional mechanisms.
>
The primary issue when I looked at this 9 months ago was the
amount of
memory required in all the routers within the routed
localized mobility
management domain. I don't have the figures handy but it
was a considerable
amount (i.e. didn't look cost-effective) for the kind of
scalability we're
interested in (Tokyo Metropolitan Area for example). A
secondary issue is
the amount of time required for the host routes to
propagate.
FWIW, the 3GPP community has some experience now with this
kind of anchor
point architecture. I have not heard of a case where failure
of the GGSN has
been an issue. Anybody on the list with 3GPP operational
experience want to
comment?
> I might have time, at least in June timeframe, but
I'll have to see based
> on how the 3 other docs look like.
>
OK. You are certainly welcome to comment on the other docs
as well, but
we're trying to finish up the PS and REQ docs at the
moment. I'm working on
resolving WG comments as we speak.
jak
_______________________________________________
netlmm mailing list
netlmm ngnet.it
https://vesuvio.ipv6.cselt.it/mailman/listinfo/netlmm
|