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

Re: [RRG] some musings on PI v. PA, and assumptions, requirements, and tradeoffs




On Jul 16, 2007, at 3:44 AM, Brian E Carpenter wrote:

On 2007-07-15 04:55, Bruce Curtis wrote:
...
Another possibility to publish the the information is to put the multiple destination addresses in each packet. Of course the cost is sending the extra bits. It seems to me that for devices in the core of the network the cost of publishing the info this way is mostly independent of the total number of routes or paths in the network but would instead be dependent on the number of destination addresses that would need to be supported per packet.

It seems to me that Six/One (and to some extent shim6) is simply
a compressed form of this - a packet contains enough info to
recover the compressed into, which is the set of addresses. The only
question is who's got a copy of the compression state.

Yes, but for similar overhead (added bits) to encap/tunneling solutions we could put the info in the packet in uncompressed form.

As you have pointed out before an architecture based on multiple destination addresses has it's own set of problems but it would allow for a future where TE could be done in the core while as you mention above the ability to do TE beyond the edge with other architectures can be quite expensive.

You could do this neatly in IPv6 with a new form of routing header
which simply lists alternate destination addresses.

    Brian

  Right, something like

http://www.watersprings.org/pub/id/draft-bagnulo-multi6-mhexthdr-00.txt

This could be implemented on hosts but would also allow middlebox solutions that strip or rewrite the header at the edge of the network so that between the edge of the network and the host there is only one destination address in the packet. (to make ACL management easier etc, also a possibility is to have one PI IP number per host that is not used for global routing but is one of the addresses in the headers plus the PA addresses etc., another option is to have the PI addresses based on location/GPS etc)

I think the property of easily supporting middlebox solutions but not requiring a middlebox (easily implemented on end hosts) is attractive.


---
Bruce Curtis                         bruce.curtis@ndsu.edu
Certified NetAnalyst II                701-231-8527
North Dakota State University


--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg