Internet Protocol v6 (IPv6)

Introduction

IP version 6 (originally referred to as Next Generation IP, or IPng) as defined in RFC 2460 has been designed to enhance the Internet protocol in the following areas:

Increase the availability of Internet address space
Enhance the ability of backbone routers to maintain large routing tables
Better support for Quality of Service (QoS) particularly with respect to real time data.
Increase security built into IP
Increase autoconfiguration facilities

Techniques such as NAT and DHCP have staved off immediate problems with address space, whereas CIDR eased the pressure on routing table sizes. Despite these tools however Internet appliances, PDAs, wireless phones, 'Ethernet to the home' and online gaming are on the increase and are not served well by these techniques. NAT for instance prevents certain applications from operating as well as creating processor load on the routing devices.

In case you are wondering why version number is 6 rather than 5, IP version 5 had already been allocated to Internet Streaming Protocol version 2. ISPv2 is detailed in RFC 1190.

Test Networks

There are two test networks that have been set up to allow testing of IPv6. The 6bone network is used primarily for testing problems, IPv4 to IPv6 transition and new protocols. It has been designed to look like a worldwide hierarchical setup with pseudo Tier I, Tier II and local site providers that communicate using BGP-4 multiprotocol extensions for IPv6 (RFC 2545).

Another network called the IPv6 Research and Education Networks (6REN) was also established to enable non-profit organisations as well as educational and research institutions to access IPv6 transit mechanisms as well as provide access to the 6bone.

Packet Format

Version - 4 bits to indicate the IP version number, in this case 6
Traffic Class/Diff Serv - 8 bits used by source nodes and intermediate routers to distinguish different priorities of IPv6 traffic. The structure of this is identical to the DiffServ field in IPv4 and functions in the same way (see QoS)
Flow Label - 20 bits used by nodes that initiate traffic flows. This is a label that tells intermediate routers to treat this flow differently from other traffic. This can be initiated by an application to override the default QoS.
Payload Length - 16 bits used to indicate the length of the packet without the header but including the Extension headers. This value is the number of octets.
Next Header - 8 bits used to indicate the header that follows the IPv6 header. This will be either an Extension header or an upper layer protocol header.
Hop Limit - 8 bits containing a value up to 255 that is decremented each time the packet crosses a router boundary. At zero the packet is discarded. The value 255 is used to prove a packet is locally sourced i.e. from the same network segment. A packet with a Hop Limit of 255 is dropped.
Source Address - 128 bits used to indicate the source address of the sending device. Compare this length of address with the 32 bits used in IPv4.

When comparing this IPv6 header with the IPv4 Header, you can see how much simpler it is. The Options field is replaced by Extension Headers that are generally processed by the end nodes rather than by every intermediate routing device. In addition, you do not have flags, ID and frag offset.

Extension Headers

An end node MUST process the extension headers in the order that they exist. To determine which extension header a node must process it looks at the preceding header, this means that extension headers must take a certain order. The IPv6 header kicks off with the first Extension header if it exists. Each Extension header has its own Next Header field which lists the next Extension header that follows it. The order is as follows:

Header	Next Header Value of the Preceding Header
Hop-by-Hop Options	0
Destination Options (examined by ALL destination nodes)	60
Routing	43
Fragment	44
Authentication	51
Encapsulating Security Payload	50
Destination Options (examined ONLY by the final destination node)	60
OSPF for IPv6	89

The available types of Extension Headers are as follows:

Hop-by-Hop Options Header

This MUST be examined by every router along the path from source to destination and HAS to be the header that follows the main IPv6 header. This extension header contains a number of TLV options that may change as the packet traverses the network if a particular flag is set. The Pad1 option pads the header with one octet whereas the PadN option pads the header with a multiple number of octets.

Routing Header

The Routing Header has a list of addresses thatare required to be visited whilst on the way to the destination. The first address to be visited is actually in the IPv6 header and this node processes the Routing Header. All the rest of the addresses are listed in the Routing Header. There is a field called the Segments Left field which indicates the number of nodes left to be visited before the destination node is reached. As each node processes the Routing Header it places the next node address in the destination address field of the IPv6 header and decrements the Segments Left value.

Fragment Header

This header is used when the path MTU is deemed to small for a particular packet. The source node either assumes the IPv6 minimum MTU of 1280 bytes or uses MTU Path Discovery to determine the smallest MTU along the path and then IT, rather than a router, has the responsibility of fragmenting the packet.

Destination Options Header

This header must be processed by the destination node. If this header comes before the Routing Header then each node in the Routing Header has to process it. If it comes after the Routing Header then only the destination node processes it.

Authentication Header

All non-changing fields in the IPv6 header and the Extension Headers are used in the authentication calculation. Changing fields such as those with options are seen as zero for the authentication calculation. This prevents fields that change along the path from invalidating the authentication.

Encapsulating Security Payload Header

This header encrypts data and places it in the data part of the header. The Tunnel Mode means that the whole IPv6 packet is encrypted whereas the Transport Mode means that just the transport layer portions such as UDP, ICMP and TCP, are encrypted (see IPsec for more detail).

IPv6 Address Format

Address Space

One of the initial drivers for the development of a new version of Internet Protocol was the limited address space afforded by the 32-bit (4 octet) address. Although over the recent years ways have been developed to get around this limitation to some extent, the address space problem still needed to be resolved in order to address the potential enormous global demand for directly reachable addresses. Choosing an existing variable address format such as the NSAP which could be as large as 160 bits, was discounted because of the extra processing overhead in dealing with the variable length address. For IPv6 a fixed address length of 128 bits was chosen. This gives a theoretical maximum number of addresses as 3.4 x 10³⁷ (approximately).

Format

IPv6 addresses are written in hexadecimal and are divided up into 8 hexadecimal fields separated by colons. An example is FEA0:D4C3:A727:0034:CF8A:2294:B39A:33FF. Each field of 4 hexadecimal characters represents 16 bits.

The problem with these addresses is that they are very long and cumbersome to type out. Most addresses however, will not be as complex as the example above. Most addresses will have a large number of zeros which have led to a number of rules that allow it to be easier to manage addresses.

You can remove leading zeros e.g. 003A can become 3A.
You cannot remove trailing zeros e.g. you cannot change 3C00 to 3C.
At least one character must be present in each field e.g. :1:A:0:
If there are multiple zero fields adjacent to each other then they can be compacted with :: e.g. :0000:0000:0000: can become ::
You cannot have multiple ::s because you would not know how many fields have been compacted by each ::
In a mixed IPv6/IPv4 environment you can represent an IPv4 address as the four rightmost fields within an IPv6 address, the format being v6:v6:v6:v6:v6:v6:v4.v4.v4.v4, note that the IPv4 addresses are 8 bits per dotted field as opposed to the 16 bits per IPv6 field.

Examples

Full Address	Shortcut	Explanation

25A0:0EA4:0004:00F5:008A:E5DD:3FCA:5BB2	25A0:EA4:4:F5:8A:E5DD:3FCA:5BB2	Leading zeros have been dropped
25A0:0000:0000:0000:0000:E5DD:3FCA:5BB2	25A0::E5DD:3FCA:5BB2	Four fields of zeros are represented by ::
E5DD:0000:0000:3FCA:0000:0000:D32C:04F3	E5DD:0:0:3FCA::D32C:04F3	Two fields of zeros have been replaced by one zero, the others by ::
0000:0000:0000:0000:0000:0000:0000:0001	::1	All the zero fields are contiguous
0000:0000:0000:0000:0000:0000:0000:0000	::

Address Structure

The address mask prefix works in the same work as the IPv4 method in that the format is /<number> where the number represents the bits that make up the prefix. The number of prefix bits is alot greater of course, plus compared with CIDR, there is much more detailed structure to the way IPv6 address space is allocated. Prefixes are allocated as follows:

Allocation	Prefix	Hexadecimal	Portion of Address Space

Reserved	0000 0000		1/256
Unassigned	0000 0001		1/256
Reserved for NSAP	0000 001		1/128
Reserved for IPX	0000 010		1/128
Unassigned	0000 011		1/128
Unassigned	0000 1		1/32
Unassigned	0001		1/16
Aggregatable Global Unicast Addresses	001	2 and 3	1/8
Unassigned	010		1/8
Unassigned	011		1/8
Unassigned	100		1/8
Unassigned	101		1/8
Unassigned	110		1/8
Unassigned	1110		1/16
Unassigned	1111 0		1/32
Unassigned	1111 10		1/64
Unassigned	1111 110		1/128
Unassigned	1111 1110 0		1/512
Link-local Unicast Addresses	1111 1110 10	FE8	1/1024
Site-local Unicast Addresses	1111 1110 11	FEC	1/1024
Multicast Addresses	1111 1111	FF	1/256

The IPv6 address format is defined in RFC 2373.

You will notice that a few of the address types have been named and highlighted. These represent the main address types (apart from the special and unassigned cases) i.e. Aggregatable Global Unicast Addresses, Link-local Unicast Addresses, Site-local Unicast Addresses and Multicast Addresses. Let us look at each type in turn.

Aggregatable Global Unicast Addresses

Defined in RFC 2374 the Aggregatable Global Unicast Addresses (AGUA) are those assigned to exchanges for allocation to users. There is a hierarchy of three levels:

Public - this topology is made up of providers that offer transit services and exchanges
Site - local to a site or organisation where no services are offered externally
Interface - the interfaces within a network

The structure of the AGUA is as follows:

Format Prefix (FP) - 001
Top-Level Aggragation Identifier (TLA ID)
Reserved (RES)
Next-Level Aggregation Identifier (NLA ID)
Site-Level Aggregation Identifier (SLA ID)
Interface Identifier

Within each topology hierarchies can be created to suit the regional IP registries or organisations. An important stipulation is that for the TLA there is a Default-free Zone where routers do not have a default route in their routing tables. These routers know every TLA-ID.

The Interface Identifier identifies an interface on a specific link and is in the EUI-64 format. The EUI-64 format is constructed by taking the interface MAC address, inserting FFFE after the OUI and then setting the Universal/Local bit to 1. e.g.

Take the MAC address 0000:0C23:5D31
Insert FFFE after the OUI to obtain 0000:0CFF:FE23:5D31
Set the U/L bit to 1 to give 0200:0CFF:FE23:5D21

Link-Local Addresses

The Link-Local Address is relevant only for the link and routers do not forward packets that contain Link-Local Addresses in the source or destination fields.

There is no TLA, SLA or any hierarchical information just zeros. The addresses start with FE8 and can be configured manually or with auto-configuration.

Site-Local Addresses

Site-Local Addresses are used strictly within the site or organisation. There are no TLA or NLA-IDs and the routers must ensure that packets carrying these addresses are not forwarded externally, in that way they are equivalent to the IPv4 private addresses.

The addresses start with FEC and the subnet ID allows you to create a hierarchy within the site network.

Anycast Addresses

The idea of Anycast is to provide a method of sending traffic to the nearest of a number of nodes. These nodes all having the same address. The Subnet-Router Anycast Address is pre-configured and all the interfaces connected to a particular link has this address assigned to them. This is a unicast address where the Interface Identifier is set to all zeros.

At the top end of the Interface IDs, 128 Anycast Identifiers have been reserved and are used to construct the Reserved Subnet Anycast Addresses. Including the eighth MSB these Anycast Identifiers run from 1000 0000 (0x80) to 1111 1111 (0xFF), as you can see the least significant bits of the Interface Identifier Field are reserved. So for an EUI-64 interface these give the Interface IDs of FDFF:FFFF:FFFF:FF80 through to FDFF:FFFF:FFFF:FFFF that can be used for Reserved Subnet Anycast Addresses.

See RFC 2526 for the reserved IPv6 subnet Anycast addresses.

Multicast Addresses

Broadcasts do not exist in IPv6 but multicasts do. Multicast addresses begin with FF and target groups of interfaces that belong to the multicast groups. You can also set up Transient multicast addresses for local use.

FF - the first 8 bits indicate that this address is a multicast address.
Flags - The first three bits must be zero whereas the fourth bit indicates whether the address is well-known (set to 0) or Transient (set to 1).
Scope - The scope of the address can be restricted to the node, the link, the site/organisation or global. This field takes the following values:
- 0 - Reserved
- 1 - Node local scope
- 2 - Link local scope
- 5 - Site local scope
- 8 - Organisation local scope
- e - Global scope
- f - Reserved
Group ID - Multicast address

Some well-known IPv6 multicast addresses are as follows:

IPv6 Multicast Address	IPv4 Multicast Address	Multicast Group

Node-Local Scope
FF01:0:0:0:0:0:0:1	224.0.0.1	All-nodes
FF01:0:0:0:0:0:0:2	224.0.0.2	All-routers
Link-Local Scope
FF02:0:0:0:0:0:0:1	224.0.0.1	All-nodes
FF02:0:0:0:0:0:0:2	224.0.0.2	All-routers
FF02:0:0:0:0:0:0:5	224.0.0.5	OSPF
FF02:0:0:0:0:0:0:6	224.0.0.6	All OSPF DRs
FF02:0:0:0:0:0:0:9	224.0.0.9	All RIP routers
FF02:0:0:0:0:0:0:D	224.0.0.13	All PIM routers
Site-Local Scope
FF05:0:0:0:0:0:0:2	224.0.0.2	All routers
Any Scope
FF0x:0:0:0:0:0:0:101	224.0.1.1	NTP
FF0x:0:0:0:0:0:0:127	224.0.1.39	cisco-rp-announce
FF0x:0:0:0:0:0:0:128	224.0.1.40	cisco-rp-discovery

Well-known IPv6 multicast addresses are defined in RFC 2375.

From the list above it can be seen that a multicast group is able to operate with multiple scopes and therefore have multiple addresses. The advantage that IPv6 has over IPv4 is that scoping is built in and defined as standard whereas with IPv4, scoping was more arbitrary (either TTL or administrative scoping). With IPv6 therefore it is easier for organisations to come together since scoped addresses means that the same address isn't used or two applications.

There is a particular type of multicast address called the Solicited-Node Address. This is used by IPv6 processes to communicate with IPv6 nodes. Apart from the Link-Local address, every unicast and anycast address has a Solicited-Node Address assigned to it. The Solicited-Node Address multicast address is created by taking the last 24 bits of the Interface ID (formed from the MAC address as described earlier)and placing it on the end of FF02:0:0:0:0:1:FF00::/104. If the Interface ID is 0200:0CFF:FE23:5D21 (as in the example earlier), then the last 24 bits form 23:5D21. Adding this to the end of FF02:0:0:0:0:1:FF00::/104 gives FF02:0:0:0:0:1:FF23:5D21 which is the Solicited-Mode Address. Because the Interface ID is unlikely to change on a particular interface, even though that interface may have a number of IPv6 addresses associated with it, only one Solicited-Mode Address is required.

Special Addresses

The Unspecified Address is 0:0:0:0:0:0:0:0 and means that no address is present e.g. on a source device that has yet to receive an address.

The Loopback Address is 0:0:0:0:0:0:0:1 and is used by a device to send packets to itself.

Other types of special addresses include IPv6 with IPv4 embedded addresses used in transition environments.