Topic: tech juniper jir prev next
tech juniper jir > 15: Introduction to IPv6
IPv6 is defined by the IETF to replace IPv4. RFC2460
The main problem with IPv4 is address space exhaustion. However, IPv6 has several other improvements over IPv4. IPv6 uses Stateless Address Autoconfiguration (SLAAC) to assign IP addresses, rather than DHCP. IPsec support is necessary in IPv6, but optional in IPv4. IPv6 has improved support for options using header extensions.
IPv6 supports more efficient routing, QoS, elimination of NAT, Network Layer security, ease of management and reduced header overhead.
The IPv6 header has a fixed length of 40 bytes. Additional options are supported via extension headers. Because the header length is fixed, the header length field from IPv4 is no longer necessary.
The fields are (length in bits);
If IPv6 needs to fragment a packet, this is accomplished with extension headers.
The Flow Label field is new in IPv6.
There are six extension headers:
Addresses are unicast, multicast or anycast.
A multicast address represents a group of IPv6 interfaces. An IPv6 packet with a multicast address as its destination travels to all interfaces in the multicast group.
An anycast address is assigned to multiple interfaces, but an IPv6 packet with an anycast address only travels to one interface, usually the closest one.
RFC4291 defines prefixes in IPv6. Special prefixes are;
| Prefix | Description |
|---|---|
::/128 |
Unspecified |
::1/128 |
The loopback |
FF00::/8 |
Multicast |
FE80::/10 |
The local link |
This is still a work in progress, however, RFC3177 defines allocation sizes for different subscriber types.
| Subscriber | Prefix |
|---|---|
| Homes | /48 |
| Small and large enterprises | /48 |
| Very large subscribers | /47 or /48 |
| Mobile networks | /64 |
| Single PC | /128 |
Unicast addresses are local or global. FE80::/10 addresses are supported on the local link only and are not routed.
Multicast addresses support sixteen different types of scope. The scope of multicast packets is identified by a four bit field in the address.
Use a common prefix (FE80::/64) on all subnets. They are only guaranteed to be unique on a single link. They are often used by routing protocols and for neighbour discovery.
| Format Prefix (3 bits) | | Global Routing Prefix (45 bits) | | Subnet Identifier (16 bits) | | Interface ID (64 bits) |
The format prefix is always set to 001 for global unicast addresses. The global routing prefix represents the highest level routed by Tier 1 ISPs and peeringpoints. It is very important that the global routing prefix bits be allocated wisely, as this will have a large impact in how traffic on the Internet is routed.
The subnet identifier is assigned by the local network engineer. The last 64 bits are reserved for the local interface ID, allowing easy configuration, via SLAAC, on the local network.
The 64 bit interface ID is based on the EUI-64 generated from the local MAC address.
To obtain the EUI-64 from a 48 bit MAC address, insert 0xFFFE between the first and last 24 bits of the MAC address, and invert the seventh bit of the new address. This is the universal/local bit. Inverting this bit signifies that the address is globally unique.
CCCCCCCC CCCCCCCC CCCCCCCC 11111111 11111111 11111110 EEEEEEEE EEEEEEEE EEEEEEEE
^ Invert this bit
Static IP addressing and DHCP are considered ‘stateful’ assignment methods.
A host determines its 64 bit EUI-64 identifier. An IPv6 router periodically sends Router Advertisement (RA) messages, or in response to an RS message. The RA message announces the existence of the router and its supported features. Router Solicitation (RS) messages are sent by hosts to discover routers.
The IPv6 router sends a prefix list in its RAs. When a host receives a prefix in an RA, it can use the prefix to configure its address.
IPv6 Neighbour Discovery (ND) combines and improves on ARP and ICMP. Defined in RFC 2461. ND is optional on IPv6 devices, but a router with ND enabled can send ND information to end devices.
RFC 3315 defines stateful DHCPv6. This may be preferred when dynamic assignment of DNS servers is required, or when the MAC address should be excluded from the IPv6 address on security grounds.
Use the show route table inet6 command to view IPv6 routes.
User the show ipv6 neighbors command to view IPv6 neighbours. Neighbours are only visible after they have been discovered through Neighbour Discovery. To force a neighbour to be discovered, use the ping command.
Multicast addresses are identified by the FF first byte. IPv6 multicast builds on IPv4 multicast with multicast groups. Some multicast packets are routed.
There are three types of multicast address;
Anycast is defined explicitly in IPv6 in RFC 2526. A packet travels to just one interface in the anycast group. Anycast addresses can be used by nodes to access one of a collection of servers providing a well known service, without having to maintain a list of addresses. They can also be used to force routing through a specific ISP by using an anycast source address.
Examples of routing protocol configurations involving IPv6.
routing-options {
rib inet6.0 {
static {
route 0::/0 {
next-hop ...;
preference 250;
}
}
}
}
show route table inet6.0 protocol static
Note that the router id is still a 32-bit value. It will be configured automatically from a numbered interface if not set explicitly. Setting the RID explicitly is recommended.
routing-options {
router-id ...;
}
protocols {
ospf3 {
area 0.0.0.0 {
interface ...;
}
}
}
Most operational commands for OSPFv3 are the same as for OSPFv2, replacing
ospf with ospf3.
The configuration of IS-IS is the same for IPv6 as for IPv4, except that IPv6 addresses are used.
interfaces {
ge-0/0/0 {
unit 0 {
family iso;
family inet6 {
address ...;
}
}
}
lo0 {
unit 0 {
family inet6 {
address .../128;
}
family iso {
address 49.001.0192.0168.0201.00;
}
}
}
}
The operational mode commands are identical.
The configuration of BGP is almost identical for IPv6 as for IPv4.
protocols {
bgp {
group int-64700 {
type internal;
local-address fec0:0:0:1001::1;
neighbor fec0:0:0:1002::1;
}
group ext-65100 {
type external;
peer-as 65100;
neighbor fec0:0:0:2005::2;
}
}
}
The same operational mode show commands are used as for IPv4.
The IPv4 to IPv6 transition method suggests using tunnels to span IPv4 networks until all intermediate routers have been upgraded.
There are four broad approaches;
A tunnel is configured using GRE.
interfaces {
gr-0/0/0 {
unit 0 {
tunnel {
source <IPV4_ADDRESS>;
destination <IPV4_ADDRESS>;
}
family inet6 {
address fec0:0:0:1000::1/126;
}
}
}
}
Routes are required to send IPv6 traffic through the tunnel.
routing-options {
rib inet6.0 {
static {
route fec0:0:0:2001::/64 next-hop gr-0/0/0.0;
}
}
static {
route ... next-hop ...;
}
}