Lo Slalom parallelo di IPv4 e IPv6 è un confronto stretto tra 27 caratteristiche dei due protocolli, che ne mette in evidenza in modo chiaro e conciso le principali somiglianze e differenze.
Una Guida “rapida” di livello CCNP da tenere a portata di mano per ogni domanda sull’argomento.
| N | Feature | IPv4 | IPv6 |
|---|---|---|---|
| 1 | N. of address bits | 32 | 128 |
| 2 | N. of addresses | 4.3 billions = 232 | 340 undecillions, i.e. billions of billions of billions of billions=2128 |
| 3 | Representation | Dotted Decimal Notation | 8 groups of 4 hex digits/each; a group is also called an hextet |
| 4 | Typical address | 192.168.0.1 | 2001:0db8:acad:0001:0000:0000:0000:1234 or 2001:db8:acad:1:0:0:0:1234 or 2001:db8:acad:1::1234 |
| 5 | CIDR – Classless Inter Domain Routing, or prefix length | From /8 to /30 for nets and subnets but also < /8 for supernets; Cisco also accepts /31 subnets, and the /32 “host mask” (an IP alone) | /48 for nets and /64 for subnets, but also < /48 for supernets, and /49 – /63 for subnet aggregation; subnets may have prefix > /64, at the nibble boundary (/68, /72…) |
| 6 | Subnet Mask | Dotted Decimal Notation, with all “1” on the left, i.e. 255.255.255.0 | Not used: network portion is the “prefix length” only, i.e. /64 |
| 7 | Classes | Class A: 1st byte 0-127 (255.0.0.0) Class B: 128-191 (255.255.0.0) Class C: 192-223 (255.255.255.0) (A, B & C are for Unicast traffic) Class D: 224-239 (Multicast) Class E: 240-255 (IANA reserved) | No Classes at all (so, no automatic summarization applies for IPv6 networks / addresses) |
| 8 | N. of subnets | It depends on the bits borrowed from the Host portion of the Major net; i.e. 172.16.3.0 255.255.255.0 has 8 bits borrowed from a Class B net, so it’s one of 28 = 256 subnets | Subnets are normally 216 (65.536) for each net, or Site prefix (/48); all the fourth hextet is dedicated to the subnets definition, by default |
| 9 | What ISPs allocate / sell to their customers | Small number of public addresses (no longer available) i.e. /32 for 1 address, /31 for 2 addresses, /30 for 4, /29 for 8, /28 for 16… Small/residential users very often only receive one dynamic public address on their WAN link interface. | Normally, ISPs allocate /48 site prefixes to their customers, regardless of their “importance”. Each customer has then up to 65.536 subnets “for free”; if no subnets are used, set fourth hextet to 0, or to 1, or to ffff, or to any! |
| 10 | N. of addresses per interface | Normally only one (but secondary addresses can seldom be used) | Normally, IPv6 interfaces have: – a Link-local address alone, for local comm. only (i.e. FE80::1/64) – a Global-unicast address, with a Link-local mate, for global comm. Global-unicast addresses can be more than one; Link-local is one |
| 11 | Size of the subnets | It depends on the bits leaved in the Host portion of the address, after subnetting was possibly done. Major Class A nets have 224 = 16,7 millions IPs; /30 only have 4 IPs | Normally, subnets have 64 bits in the IID-Interface ID portion of the address, the rightmost half of it. Thus, IPv6 subnets allow for 264 = 18.5 billions of billions of IPs |
| 12 | Scopes of addresses | Unicast (Classes A, B & C) Multicast (Class D) Broadcast (directed 192.168.0.255 or limited 255.255.255.255) | Unicast (different types –> see) Multicast (FF00::/8); includes the “broadcast concept” (= FF02::1) Anycast (these are Global-unicast assigned to different devices) |
| 13 | Special and public addresses | 0.0.0.0/32 = unspecified IPv4 0.0.0.0/8 = reserved 127.0.0.0/8 = loopback / local host (16 million addresses… wasted) 10.0.0.0/8 = private Class A network 172.16.0.0/12 = private Class B network 192.168.0.0/16 = private Class C network 169.254.0.0/16 = APIPA-Automatic Private IP Addressing / Link-local All other A-B-C Class are public | ::/128 = unspecified IPv6 ::1/128 = loopback / local host (only 1 address of the… infinite) FC00::/7 = Unique-local (ULA), routable only within their Site FE80::/10 (RFC 4291: FE80::/64) = Link-local, non routable, only for local communication on a link FF00::/8 = Multicast, of 5 scopes 2000::/3 = public Global-unicast (small parts of 2000::/4 assigned, mostly from 2001::/16) |
| 14 | Documentation addresses | 192.0.2.0/24 209.165.200.0/24 …201.0 …202.0 and few others (seldom used) | 2001:db8::/32, supplied by APNIC; third hextet is typically “acad”, “beef”, “cafe”, “feed”… |
| 15 | NAT | Different uses of NAT allow: – Client with private addresses to access the Web (dynamic + PAT) – DMZ Servers to be reachable from the Web on public addresses – merging of two networks with similar private addressing, etc. | No NAT is used for IPv6: Global-unicast addresses are public. NAT64 is used for IPv4 – IPv6 coexistence in the short term, by replacement of Packet headers |
| 16 | Layer 3 to Layer 2 address resolution protocol | ARP – A special Packet (Ethertype code 0x806) is sent to L2 broadcast (ff-ff-ff-ff-ff-ff), requesting the MAC address of the local node with a given IP | ND – Neighbor Discovery: a NS-Neighbor Solicitation ICMPv6 message (code 135) is sent to the target node, requesting a NA-Neighbor Advertisement (code 136) from it, with its MAC address |
| 17 | ICMP and PING | The ICMPv4-Internet Control Message Protocol v4 includes up to 15 different message types (from 0 to 18, with holes); the most used types are: 0=Echo Reply, 3=Destination unreachable (with subcodes), 8=Echo Request and 11=Time Exceeded. PING = 8 –> 0 | The ICMPv6 protocol includes many messages for functions similar to the ICMPv4 ones, plus the ND-Neighbor Discovery suite described above, including also the RS-Router Solicitation (code 133), RA-Router Advertisement (code 134) and Redirect (c. 137). PING = 135 –> 136 |
| 18 | Static address configuration methods | Only one method is available: just configure the address and its Subnet Mask: that’s all, folks! | Three static methods are available to setup a Global-unicast address: – fully static: assign all 128 bits – EUI-64: assign the /64 prefix and add eui-64: the IID is derived from a local MAC (see A1) or from the fully static Link-local – random: only on Windows PCs |
| 19 | How to enable the routing globally on Cisco IOS | IPv4 is already enabled by default on Cisco Routers and Multilayer Switches | To turn a Router or a Multilayer Switch into a dual-stack device, use the command: D(config)#ipv6 unicast-routing |
| 20 | How to enable an interface on the protocol on Cisco IOS | IPv4 is always enabled by default on Cisco Routers and Multilayer Switches routed interfaces; just add an IP and no shut to the interface | By adding the command: D(config)#ipv6 enable to an interface, it generates its EUI-64 Link-local address |
| 21 | Dynamic address configuration methods | After ICMP, RARP and BOOTP made the history, today only the DHCP-Dynamic Host Configuration Protocol is available (RFC 2131 & 2132 for optional fields): Cl.–>Se. DHCP Discover (L2 bro) Se.–>Cl. DHCP Offer (L2 unicast) Cl.–>Se. DHCP Request (L2 bro) Se.–>Cl. DHCP Ack (L2 unicast). Other messages are defined, i.e. Nak, Decline, Release and Inform. DHCP can supply > 30 parameters | Again, three dynamic methods: – SLAAC-StateLess Address AutoConfigura-tion: the node finds a Router on its link (–> FF02::2) and receives the prefix and D.G. (Router’s Link-local) + EUI-64 – Stateless DHCPv6: the same as above, but a DHCPv6 Server can add DNS, domain name, NTP, etc – Stateful DHCPv6: the Router delegates the duty to a DHCPv6 Server (sometimes… to itself!) |
| 22 | How to get a dynamic address under Cisco IOS | D(config-if)#ip address dhcp | Referring to the three methods: – D(config-if)#ipv6 address autoconfig for methods 1 & 2 – D(config-if)#ipv6 address dhcp for Stateful DHCPv6 method |
| 23 | How to route DHCP requests to a remote (helper) Server | D(config-if)#ip helper-address ipv4-address-of-Server on the interface receiving the DHCP Discover messages | D(config-if)#ipv6 relay destination ipv6-address-of-Server on the interface receiving the DHCPv6 Solicit messages |
| 24 | D.V. routing protocols | RIPv1, RIPv2, IGRP, EIGRP | RIPng, EIGRP for IPv6 EIGRP is activated by no shut. Networks are supplied to protocol in the D(config-if)# mode |
| 25 | L.S. routing protocols | OSPFv2, IS-IS | OSPFv3 Networks are supplied to protocol in the D(config-if)# mode |
| 26 | Packet Header | It is normally a 20 bytes header, with 12 fields (see A2), namely: Vers.=4, IHL=5, ToS, Pkt length, Identif + Flags + Pkt offset (for Packet fragmentation), TTL, Protocol (i.e. 1=ICMP, 6=TCP, 17= UDP, 41=IPv6), Header checksum, Source address, Destination address | It is normally a 40 bytes header with 8 fields (see A3), namely: Vers.=6, Traffic class, Flow label, Payload length, Next header, Hop limit, Source address, Destination address. Additional Extension Headers can follow, for fragmentation, VPN… |
| 27 | Ethertype | The 16 bits code used by Layer 2 protocols to identify IPv4 is 0x800 = 0000 1000 0000 0000 | The 16 bits code used by Layer 2 protocols to identify IPv6 is 0x86DD = 1000 0110 1101 1101 |
Annexes
A1 – EUI-64 method to generate a unique IID for IPv6 addresses
A2 – IPv4 Packet Header
A3 – IPv6 Packet Header
