IPv4 vs IPv6: addressing, SLAAC, dual stack

The short answer, then the mechanics

IPv4 uses 32-bit addresses, IPv6 uses 128-bit addresses, and nearly everything else that exams ask follows from that and from three design changes: IPv6 has no broadcast (multicast replaces it), hosts can configure themselves with SLAAC (stateless address autoconfiguration) from router advertisements, and every interface carries a link-local address (fe80::/10) that is used for neighbor discovery and routing next hops. The decision rule for scenario questions: for “a host must get an address with no DHCP server” think SLAAC, for “which address is valid only on this link” think link-local, and for “run both protocols side by side during migration” think dual stack. The addressing architecture is RFC 4291, and the address length, link-local, multicast and no-broadcast facts below come from it.

Layer 1: reading and shortening an address

An IPv6 address is written as eight groups of four hexadecimal digits separated by colons. Two rules shorten it, and they are the source of many exam questions:

1. Leading zeros in a group can be dropped (0db8 becomes db8, 0000 becomes 0).
2. One run of consecutive all-zero groups can be replaced by ::. Only one, because with two a reader could not tell how many groups each stands for.

2001:0db8:0000:0000:0000:ff00:0042:8329 becomes 2001:db8::ff00:42:8329. The reverse also works as a check: count the groups that are present, subtract from eight, and that many zero groups sit where the :: is. A prefix length works as in IPv4: 2001:db8:1:1::/64 means the first 64 bits are the network prefix. A normal LAN subnet is a /64, because SLAAC and the standard interface identifier assume 64 bits of host part. The prefix plus interface identifier must add up to 128 bits, and RFC 4862 says a host ignores an advertised prefix for SLAAC if they do not.

Layer 2: the address types

Anycast has no prefix of its own: an anycast address is taken from unicast space and looks identical to a unicast address, and the same address is configured on several nodes so the nearest one answers. The 2000::/3 range is how global unicast space is currently allocated; RFC 4291 itself defines global unicast as everything not claimed by the other types.

Layer 3: how a host gets its address

SLAAC. The host first builds a link-local address from fe80:: plus an interface identifier, then listens for Router Advertisements (RAs) from routers, or asks with a Router Solicitation. Each RA carries a prefix; the host appends its 64-bit interface identifier and has a global address. Before using any address the host runs Duplicate Address Detection with a Neighbor Solicitation. No server tracks leases.

DHCPv6. Stateful DHCPv6 hands out addresses and keeps leases like DHCPv4. Stateless DHCPv6 is a combination: the host builds its address with SLAAC and asks DHCPv6 only for extras such as DNS servers. The RA’s flags tell the host which to use (the managed and other-configuration bits are in RFC 4861). Note that DHCPv6 does not supply the default gateway; the router’s RA does.

Static. Configured by hand, still subject to duplicate address detection.

The interface identifier can be derived from the MAC address by modified EUI-64, the form CCNA 200-301 tests by name and the one a Cisco router uses with ipv6 address <prefix>/64 eui-64. Because it exposes the MAC address, RFC 8064 recommends stable, opaque identifiers for SLAAC instead, and current operating systems generate random-looking identifiers by default.

Worked example: EUI-64

A host interface has MAC 00:1A:2B:3C:4D:5E and the RA advertises 2001:db8:10:20::/64.

1. Split the MAC in the middle: 00:1A:2B and 3C:4D:5E.
2. Insert FF:FE between them: 00:1A:2B:FF:FE:3C:4D:5E.
3. Flip the seventh bit of the first byte (the universal/local bit). 00 is 0000 0000; flipping the bit worth 2 gives 02.
4. Write as four groups: 021a:2bff:fe3c:4d5e.
5. Join to the prefix: 2001:db8:10:20:21a:2bff:fe3c:4d5e (leading zero dropped).

What changed from IPv4 besides length

Dual stack and the other migration options

Dual stack runs IPv4 and IPv6 on the same devices and interfaces, so each host has both addresses and uses whichever the destination supports. Tunneling carries IPv6 packets inside IPv4 across a network that does not route IPv6 yet. Translation converts between the two. RFC 4213 describes dual stack and configured tunnels. For routing, dual stack means two sets of everything: ipv6 unicast-routing must be enabled on a Cisco router, and static routes and OSPF exist per protocol (OSPFv3 for IPv6; see OSPF explained).

Which exams test this

How exam questions are built around the topic

Expect one of four shapes. Shorten or validate an address: the trap is a second ::, or a group with five digits. Identify the type from the prefix: fe80 is link-local, fd or fc is unique local, ff is multicast, 2 or 3 as the first digit is global unicast. Build the EUI-64 address: forgetting to flip the seventh bit is the standard wrong answer. Autoconfiguration scenarios: “no DHCP server, clients still have global addresses” means SLAAC and router advertisements; “clients have an address but no DNS server” points to stateless DHCPv6. Constraint words that decide: “link”, “site”, “internet”, “stateless”, “nearest”.

A common confusion is thinking IPv6 needs a /64 only by convention. For SLAAC it is effectively required by the 64-bit interface identifier. For subnetting arithmetic, the IPv4 method in subnetting step by step carries over, but exams rarely ask for IPv6 host counts. Addressing rides on the layers of the OSI vs TCP/IP models, and NAT questions are covered in NAT types.

Practice questions

Go deeper

• RFC 4291, IP Version 6 Addressing Architecture
• RFC 4862, IPv6 Stateless Address Autoconfiguration
• RFC 8200, IPv6 Specification
• RFC 4193, Unique Local IPv6 Unicast Addresses
• Related on this site: DHCP and DNS, the CCNA study plan and how to study for certification exams.