Hi Qi, BR1 and BR2 have an same IPv6 anycast address. CE can send out BR anycast IPv6 packet where IPv4 packet is encapsulated. So, when BR1 is going down, BR2 can receive the packet from CE.
From CE perspective, as Ole mentions, CE needs to maintain a default route toward to MAP in order to forward IPv4 packet over MAP domain. At CE, IPv4 packet is encapsulated in IPv6 and then IPv6 packet is reached to BR. If this IPv6 packet is terminated on the tunnel end-point address at BR, MAP in BR can process this packet. On the other hand, when BR receives IPv4 packet from IPv4 network, BR needs to forward the IPv4 packet over MAP domain. For this, BR needs to maintain an IPv4 route corresponding to Rule IPv4 prefix. For example, BR has a following IPv4 route. 1.1.0.0/16 -> MAP0 In the above example, I assume 1.1.0.0/16 is used for Rule IPv4 prefix. Thanks, Tetsuya Murakami On 2012/03/02, at 0:52, Qi Sun wrote: > Hi, > Thank you for your reply~ > Here is my understanding and i'm not sure if it's right. > When BR1 is down and the IPv4-encapsulated packet comes to BR2, BR2 can > handle the packet properly using the same MAP rules with CE1 because they are > in the same domain. BR2 doesn't have to handle the source IPv4 address, > saying 1.1.1.1. Right? > > Looking forward to ur reply~ > > Best Regards, > > Qi Sun > > From: Ole Tr鴄n > Date: 2012-03-02 03:43 > To: Cameron Byrne > CC: softwires WG; sunqi.csnet.thu > Subject: Re: [Softwires] Comments on MAP-E draft > Cameron, > > >>> What is the expected behavior when that address becomes unavailable? > >> > >> reliability is achieved through anycast. there are multiple BRs > >> advertising the same address into the IGP. > >> did that address your question? > >> > > > > Sorry, i have not really been able to find this out on my own. > > > > But, i thought there was tight coupling of the BR address resources > > and CPE address assignments. > > > > Meaning BR1 has for me (the subscriber) the public address 1.1.1.1 > > with ports 2000-3000 and that is communicated to CPE1 > > > > BR1 goes has a fault, goes off line. > > > > BR2 is now the closest anycast path to CPE1 > > > > My CPE is sending packets to BR2 (as anycast) assuming BR2 owns > > 1.1.1.1 and port set 2000-3000. > > the BR doesn't really have any port sets, although it could. > the anycast address is an IPv6 address. > > > > > But, how does BR2 know anything about 1.1.1.1? Does BR2 now announce > > in BGP to the world it owns 1.1.1.1? > > > > Is that going to work? If so, where is it documented? > > > > Or, does the CPE have to reboot and get some new config assignment > > based on BR2 now being the closest? > > > > How does this oscillating load (BR failure, shortest path changes, ... > > ) associated with anycast bode for the IPv4 address reserves ? Seems > > like the public IPv4 address pool on each BR would have to be > > over-provisioned to handle some N*X level of load. > > > > Sorry if i am missing something obvious. > > sounds like we need to make this a little clearer in the draft. > > it isn't quite like you describe. > the BR is used for out of domain traffic. > other traffic is 'on-link'. > > e.g. the CE has a default route: > 0.0.0.0/0 -> 2001::1, MAP0. > > 2001::1 is advertised from every BR. > > cheers, > Ole >
_______________________________________________ Softwires mailing list [email protected] https://www.ietf.org/mailman/listinfo/softwires
