IPv6 Subnetting

Question

Hello,
So far, as i understood from different web searches and resource materials, the Internet Registry (ARIN, for example) provides a /32 address block to Top ISP. The Top ISP provides a /48 to customers or to normal ISPs (Lower ISPs). Finally, the customer breaks the /48 into subnet using the 16 bits available. (2^16=65535 /64 subnets). Here my question comes in, How is subnetting performed by this way? ... if 65535 subnets to be used, how can we categorize them into Class A, B and C like in IPv4? .... About hosts, i understand that the low-order 64 bits in the ipv6 address is used for hosts but also how do i know which hosts to assign to which subnets? ... and because VLSM is not necessary in IPv6, does it mean that the Classless Routing Protocols concept is gone and IPv6 will be Classful?
Someone kindly shed some lights on how IPv6 subnetting is performed step-by-step.
Your response is highly appreciated.
Regards,
AM

Answer 1

The first thing with IPv6 is to forget everything you learnt about IPv4 conservation. To put it crudely be wasteful and do not be concerned that you have 1.8x10^18 useable host addresses in a /64 and are only going to use it to address a point to point link or use it for a VLAN allocation for 15 servers.
(I am being overly simplistic here to illustrate a point, there are some subtleties, though in essence go with this)
There are also no classes, no network addresses, no broadcast addresses, no wild bit masks. Everything is CIDR, all addresses are useable.
So in the example of getting a /48, you get 65536 x /64 prefixes (2^16bits of subnetting)
Then just start assigning /64s everywhere you would assign any sized subnet in IPv4 (with the exception of loopbacks)
So from the 1st /64 sequentially number all loopbacks for devices using /128 (e.g. prefix::1, ::2, ::3, ::4 etc)
then your first point to point link gets the next /64, and the next point to point link gets the next /64 and so on, then your first VLAN gets the next /64 and so on.
So if you have 20 devices, 34 point to point links and 12 VLANs in your network you would ultimately be using 47 x /64 prefixes from your /48 and keep the rest (65489 x /64s) for later use.
That's the subnets taken care of. Hosts are in general addressed in the last 64 bits if the prefix. This can be done in one of three ways:
Manually configured (e.g. prefix::3/64, prefix::7/64, prefix::AB/64, prefix::cafe:f00d/64),
Using DHCPv6, or
Using SLAAAC (Stateless Address Auto Configuration) where the host generates its own IPv6 address using its MAC address and padding it out to 64 bits. (Do searches in these terms for more info)
String it all together using the IGP of your choice, and that is pretty much it.
I hope this helps