Testing: http://test-ipv6.com/

Understanding: http://fr.wikipedia.org/wiki/IPv6


with Windows: http://www.youtube.com/watch?v=0RvJG96iwBg

with Linux: https://www.linux.com/learn/tutorials/428331-ipv6-crash-course-for-linux

=> doc from ovh: http://guides.ovh.com/Ipv4Ipv6#link6

My feelings and IPV6 101:

IPV4 has 2*32 addresses = 4,22 billions  ; IPV6 has 2*128 = 3,4 10*38 addresses (soit 667 millions de milliards d’addresses par mm2)!

IPV6 is constitued of 8 blocs of 4 Hex => FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF

the red part: represents the number of Networks ; the blue part represents the number of hosts (computers, sensors, smartphone…) is also noted /64. The red part is managed by the operators/ISP…

in terms of networks: we have in general the networks starting by the first bloc of bytes:

FE80:  correspond to link-local == equivalent to private addresses in IPV4. Is not routable. >>>> a link-local address is automatically used by default everywhere for each IPV6 compliant device

FC00: correspond to unique-local (ULA) == equivalent to 172.x or 192.x or 10.x in IPV4. Is not routable.

2000: correspond to global == internet addresses in IPV4. Is routable. Those addresses are distributes by the ISP / Registrars

The IPV6 header is simpler than IPV4 header. But in reality it is not very important, because in terms of performance => due to improvements of the hardware, nowadays it is not an issue.

No modification on TCP transport protocol. ICMP modified to support ipv6. In terms of routing protocol : RIPv2, OSPFv3, BGP4+ supports ipv6 now. The most important to understand in IPV6 are the new protocols in charge of link-operations (all the protocols implied by the IPv6 addresses delivery, management and controls): by analogy it is equivalent to the ipv4 protocols : ARP+ICMP+DHCP

The most important in link-operations are the two parts:

  • Neighbors discovery protocol also called ND ( constitued by RS,RA and NS,NA traffic)
  • Duplicate address detection protocol also called DAD

Methods to deploy IPV6 in enterprise: Link-Local only is enough, but not routable. In general the companies requiring access to internet uses ULA+global IPV6 subnets. In terms of cohabitation with ipv4, the prefered method is to use devices (routers/hosts) which supports the dual stack IPv4/IPv5, the second well-known method is the Tuneling (there are a dozen of tuneling protocols). The third method (but not recommended due to perf issues) is the Translation (ipv4>ipv6)

tip1: IPV6 is more secure ? => false ; but ipv6 harder to discover (due to larger ipv6 range “in general a /64 attributed”).

tip2: IPV6 requires IPSEC? => false ; it is not mandatory at all