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

Re: [Ans-research] IRTF ANS Meeting Announcement

On Sunday, June 1, 2003, at 02:09  AM, Pekka Savola wrote:

On Sat, 31 May 2003, S Woodside wrote:

I'm curious if a large adhoc network would have special multihoming
requirements that multi6 might consider? Put together, scalable adhoc
networks with the ability to multihome in a decentralized manner would
be a very desirable technology for building flexible wireless networks
for last-mile (urban) / last 15 miles (rural) access.

Perhaps it would be best to first try to figure out how single-homed
internet connectivity would work :-).
This is done, fortunately, for a small mesh at least. Unfortunately, the system uses centralized management to configure which (single) gateway each node uses (the system is meshAP, the management is called wiana). I don't think management is necessary though. AODV should be capable of handling single-homing by having the gateway node route all of the external packets. The existing adhoc protocol would handle the rest.

But really, it's still a question on how you delegate the information
about prefixes and choose your subnets.

For the multihomed case with multiple PA prefixes, I guess this would mean
that every MANET node would have a fixed, unique-in-the-MANET site
identifier (or multiple of them). This could be used for delegating
addresses and ensuring multiple nodes don't get weird ideas which parts of
/48's they could use.
Starting with a simplistic analysis, the nodes would not know the MANET is multihomed. The gateways could each all advertise the external network. It's possible that a MANEt protocol would be able to handle this, if it selects (for example) the first path that returns a route. In that case nodes would probably route to the outside through the closest gateway node.

Or, the gateway nodes could discover each other and set up tunnels over the MANET to coordinate which sections of the external network they will each advertise in the MANET. That will add overhead and complexity.

Or, I think what you're saying is, that each NODE would know about all of the gateways, and which prefix they have, and make its own decisions about which to use. That would require some additional information in the MANET protocol to essentially separate identity and locator, to allow multiple gateway nodes to advertise the same identities, but with different locators.

(if I'm not totally confused. I just read RFC 2101)

simon




Begin forwarded message:




From: Elizabeth Belding-Royer <ebelding@cs.ucsb.edu>
Date: Thu May 29, 2003  8:11:49  PM America/Montreal
To: ans-research@www1.ietf.org, manet@ietf.org
Cc: corson@flarion.com, ebelding@cs.ucsb.edu
Subject: [Ans-research] IRTF ANS Meeting Announcement

Hi,

The first meeting of the new IRTF Ad hoc Network Scalability (ANS)
research group will occur this Sunday at the Mobihoc conference
hotel.  Included is the agenda for this meeting.  This is a somewhat
fluid agenda as we are still finalizing the speakers.  However,
the meeting will definitely start at 7pm Sunday evening.  We look
forward to seeing many of you at the meeting.

Scott and Elizabeth

*****************************************************************


Date: Sunday, June 1, 2003
Time: 7-10pm
Location: Sheraton Barcelo Hotel (Mobihoc hotel)
  I don't have the exact room at the moment - an email will be
  sent later with that information.


Scott and Elizabeth (10 min)
  - Welcome
  - Overview of research group charter, objectives

Scott and Elizabeth (30 min)
  - Recap of mailing list discussion on direction of research group
  - Group discussion on direction/objectives

Presentations

 * Zygmunt Haas (20 min)
   - Hybridization and Adaptivity as Means to Scalability
   - Scalability Bounds in Wireless Networks

 * Joe Macker (15 min)
   - Link state scaling approaches

 * Mario Gerla (15 min)
   - Routing approaches and scalability metrics.

 * Radha Poovendran (15 min)
   - Greedy Perimeter Routing

 * Christian Bettstetter (15 min)
   - Interconnection of Ad hoc Networks to IPv6 Networks

Open Forum for Discussions (1 hour)







--
www.simonwoodside.com -- 99% Devil, 1% Angel