Re: ND (NUD) on ISATAP links
Vladislav Yasevich <[email protected]> Thu, 17 Oct 2002 17:54:51 -0400
| Newsgroups | gmane.ietf.ngtrans |
|---|---|
| Organization | Hewlett Packard |
| Message-ID | <[email protected]> |
Jason Jason Goldschmidt wrote: > > > As Karen pointed out, performing NUD between ISATAP hosts and each > router in its PRL will impose scalability problems for ISATAP. If using > the default REACHABLE_TIME value of 30 seconds, this will impose a large > strain on ISATAP's routers for sites with 1000+ ISATAP hosts. Obviously > REACHABLE_TIME could be increased, but then it is not clear what value > implementing NUD would buy over simply decreasing the minimum value for > router lifetime (and in hand decreasing MinRouterSolicitInterval). NUD is not performed between hosts and "each router in its PRL". It is performed between a host and a (one) router from the PRL that is currently being used. The router that will be selected by the host will depend on the traffic patterns from that host and any destination cache entries that may exist. The reason for NUD is that the IPv6 address of the router may not be reachable even though the IPv4 address is reachable. This is exaclty what NUD is used for. The scalability problem you mention only happens when there are very few routers serving a lot of hosts. In that case, a particular router is more likely to be picked as a next hop and NUD will be run after the AdvReachableTime has expired. The consuption of resources on that ISATAP router will roughly equivalent to a normal IPv6 router in a similar configuration (+ the extra is a tunnel overhead) so I don't see a large problem with scalability. > > Thus I'm not convinced implementing NUD to prevent black-holing is a > viable solution. > > Can you please explain the benefits of having an ISATAP host connect to > multiple ISATAP routers? If the goal is for redundancy, this might be a > hard goal to achieve if the above problem can not be solved. It may be redundancy. It may be multihoming (you have 2 different ISATAP prefixes). I think the authors of the draft can comment more since multiple routers are explicitely described in the draft. -vlad > > -Jason > -- ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Vladislav Yasevich Tru64 UNIX - IPv6 Project Lead Hewlett Packard Tel: (603) 884-1079 Nashua, NH 03062 ZKO3-3/T07