Your IP Address is: 38.103.63.61
[moonv6] about multihoming testing
From: Marc Blanchet (Marc.Blanchet@hexago.com)
Date: 02/27/04
- Next message: Green, David B RDECOM CERDEC STCD SRI: "RE: [moonv6] conference call dial-in info"
- Previous message: Bound, Jim: "RE: [moonv6] Moonv6 Vision"
- Next in thread: CATANZARITI Sergio / FTR&D / US: "RE: [moonv6] about multihoming testing"
- Maybe reply: CATANZARITI Sergio / FTR&D / US: "RE: [moonv6] about multihoming testing"
moonv6 post from Marc Blanchet <Marc.Blanchet@hexago.com>
As requested by Ben during this week conf call, here is some ideas on a
possible test plan items for multihoming.
Out of the things we _could_ test at the moment are the following. Here I'm not discussing pros and cons of approaches.
Possible multihoming solutions that could be tested now:
- ipv4-like multihoming: customer have independent address space, announced
by more than one ISP
- one /48 prefix by each provider to the site: multiple prefixes announced
on the site. default address selection algorithm on the host is taking over.
3484 Default Address Selection for Internet Protocol version 6 (IPv6).
R. Draves. February 2003. (Format: TXT=55076 bytes) (Status: PROPOSED
STANDARD)
This RFC is implemented by many IPv6 OS stacks. Test cases could be done on how hosts are behaving when multiple prefixes are advertised (so configured) on the host, and the destination is one or the other provider prefix, a very near prefix, etc... Also, how the address selection default is modified by the table to have a higher priority on one provider (that could be considered as the "main" provider for the site). Multiple test cases can be written for default address selection, and testing of host stack implementations will be useful (as well as router advertisements with multiple prefixes, different lifetimes, etc...)
- cross-tunnels with providers at edge.
3178 IPv6 Multihoming Support at Site Exit Routers. J. Hagino, H.
Snyder. October 2001. (Format: TXT=24453 bytes) (Status:
INFORMATIONAL)
One obvious test case would implement this solution and see how the routing works.
my 2 cents.
Marc.
This archive was generated by hypermail 2.1.7 : 12/01/06 EST