IPv6 Track Interface Delegation size question

[MageMinds]

I managed to make my IPv6 setup to work when I select /56 in the delegation of my WAN and the LAN and my other two VLAN they receive /64. The /64 is the right size that my provider gives me, but why do I have to ask a delegation of /56 to make it work, If I choose a delegation of /64 in the WAN config then the LAN and the VLAN try to setup a /56… Looks like things are reversed in the config somehow...

WAN Configuration

Interface status

[kpa]

What they are delegating to you via their routes is a /56 and it breaks down to a maximum of 256 different /64s that can be used on your different LAN networks. Your ISP can't route each of the /64s separately because that would unmanageable.

[gerdesj]

@kpa:

What they are delegating to you via their routes is a /56 and it breaks down to a maximum of 256 different /64s that can be used on your different LAN networks. Your ISP can't route each of the /64s separately because that would unmanageable.

To clarify a bit:

One of the major design goals of IPv6 was to simplify the global routing tables (*).  The original RFCs recommended a /48 per subscriber which is rather a lot.  Many ISPs have reduced that to /56 and some to /64.

Now the smallest IPv6 subnet size is /64 - yes it is huge but that is the design and it works.  Do not try to divide a /64 up - things will go funky!

So, you have 256 x /64 allocated to you.  It is fine to only use one and ignore the rest.  It may seem wasteful but it isn't really, because there are rather a lot of /56s available and if you decide to create say the following VLANs:

LAN (already done)
SERVER
wLAN
wPHONE
wPUBLIC
THINGS
MANAGEMENT

… and maybe a few more, then you can.

Heck, you could provide VLANs/subnets for your neighbours by splitting your 256 VLANs into a site ID and a subnet ID for 16 x 16 of each.  A /48 gives a magic number of 256 x 256 which maps rather nicely onto IPv4 which is probably why it was chosen.

You may find it amusing to note that a point to point link will often have a separate /64 eg your PPPoE/A link will probably have 2^64 addresses available but only two are used.  Is that more wasteful than a maxed out collision domain of say 2000 devices: 2/(2^26) is a tiny number and so is 2000/(2^64).  Both are smaller than the probability of winning a national lottery!

Anyway, the whole point of the design is that you should not be constrained by addressing and aggregation of routes should be easy (**).

Cheers
Jon

(*) The IPv4 address tables are absolutely huge and fragmented beyond belief as people scrabble for chunks of the remainder and try to reclaim little bits
(**) PI - Provider Independent addressing will bugger that up eventually

[kpa]

Yeah, the PPPoE example is exactly the same situation that you have with a tunnel provider such as HE or SixXS. There has to be a separate tunnel network (sometimes called transfer network) with an address space that does not overlap with any of the other subnets used on your systems, otherwise you just can not route anything. On such tunnel network a /64 is used but only two addresses are ever actually used because link is point-to-point with just two peers.