IPv6: Internet Gets Bigger Bus

Today, in a world driven by the Internet, IP has become the de facto protocol

for most of the data communication. The existing IPv4 address space has served

the world well for over a decade, and today, thanks to its widespread success,

we are at the risk of exhausting it. Anticipating this, the IETF has drafted a

comprehensive set of specifications for the next generation protocol or address

space called IPv6.

Apart from offering a larger address space, IPv6 also has more inherent

benefits like stateless auto-configuration and in-built security, which

facilitate the next generation network applications like mobile. These have been

built in while redesigning IP.

Ipv6: An Overview

IPv6 was approved as a standard by IETF in the year 1994. Multiple working

groups simultaneously focused on the different areas that are involved in the

protocol redesign– addressing architectures, routing protocols, security, QoS,

etc. The addressing model of IPv6 and the associate routing protocols are

briefed in the following sections.

Addressing architecture

IPv6 addresses are 128 bits in length, and hence it can

support, 3.4x1038 individual addresses. In contrast, IPv4 could potentially

support a maximum of 4.2x109 individual addresses, which was considered enormous

at the time IP was evolved. IPv6 has been designed up from the ground level to

support a flexible and an efficient global routing hierarchy.

Although there are a number of allocation schemes,

aggregation-based hierarchy is preferred because it combines the power of

geographic and provider allocation schemes. In aggregation-based allocation, the

top of the hierarchy would be top level aggregators (TLAs), which are the public

transit points where long-haul providers and large telcos establish peer

connections.

With the above specification, the first three bits in the

address represent the type of address (unicast, multicast, anycast, etc), while

the next 13 bits are reserved for the TLAs. This can be represented as follows :

The 32 bits after the TLA represent next level aggregators (NLA).

This field can be further subdivided as per the NLA’s hierarchy requirements.

The last two fields represent SLA and Interface ID. SLA can

be assigned to huge enterprise networks who can address their networks with as

many as 65,535 subnets. Typically, an Interface ID will be derived from the

physical MAC address of the network element.

For example, one of the class ‘A’ service providers in

India, would allocate a TLA address, and one of their large customers would

slice the SLA into multiple hierarchies to represent the functional or

geographical allocation, and then an individual PC would take the interface ID

from its NIC card.

IPv6 routing protocols

Routing protocols are an essential part of the IP

infrastructure, and RIP, OSPF, IS-IS and BGP, are all being re-designed to

support IPv6 natively.

Interior Gateway Protocols

RIPng is the IPv6 variant of the routing information protocol

(RIP), which is in common use as the IGP for small- to medium-sized computer

networks. As the first and the simplest routing protocol standardized for IPv6,

RIPng will probably see wide adaptation during the initial phases of IPv6

implementation.

In case of large networks, where link state protocols are

used, the choice for the deployment is OSPFv3, which has been designed for IPv6.

Although OSPFv3 retains most of the algorithms from OSPFv2, certain changes have

been called for due to the changes in the protocol semantics between IPv4 and

IPv6, and also to handle the large address size of IPv6. RFC 2740 goes into more

details on OSPFv3.

Exterior Gateway Protocol

Routers currently use Version 4 of the Border Gateway

Protocol (BGP) for routing between autonomous systems. RFC 2283 defines

multi-protocol extensions to BGP that allows it to carry information of networks

other than its native IPv4, including IPv6 prefixes. This will provide the

routing of IPv6 networks over IPv4 clouds.

Transition Mechanisms

Many transition methods have been discussed for the IPv6

deployments and some of them are detailed out in this section.

Dual stacking

This calls for the configuration of both the IPv4 and IPv6

protocol stacks in all the hosts. Most of the operating systems support this

feature. Though dual stacking is possible in the end-stations and network

equipment, it is very difficult to maintain two different address space and

route on an Internet wide scale. Hence, this can be considered as a solution

only at an enterprise level and may not be for the global Internet.

Transitional address structures

Two transitional address structures have been defined–IPv4-compatible

IPv6 address and IPv4-mapped IPv6 address. In both of these, the lower 32 bits

of IPv6 are mapped with IPv4, while the higher order bits are padded with some

recognizable values.

The first address structure is used when two IPv6 networks

communicate over an IPv4 cloud. The second is used when the two IPv4 networks

communicate over an IPv6 cloud.

RFC 3056 and 2766 specification

RFC 3056 provides the specifications for connecting the IPv6

domains over IPv4 clouds and this is generally referred to as 6to4. An

end-to-end IPv6 communication is established over tunnels. The end points of the

tunnels are identified using a reserved 6to4 prefix, where the NLA field of IPv6

represents the globally unique IPv4 address for the site.

RFC 2766 is also called Network Address Translation-Protocol

Translation (NAT-PT). This defines a translation procedure from IPv6 to IPv4

address, including the packet header format and vice versa. This will be

deployed at the interconnection points and this method eliminates the need for

dual stack at the network hosts level.

The RFC2766 and 3506 will provide a flexible and

comprehensive transition from IPv4 to IPv6.

IPv6 and Mobile

Mobile usage has rocketed globally in the last few years. The 3G

specifications for wireless communication calls for packetization of mobile

voice traffic and also re-designing the wireless networks to support high-speed

data communication. Today, mobile has become a major thrust for moving to Ipv6,

as the number of addresses required by the millions of mobile users make it

unfeasible to use IPv4.

With the existing routing mechanisms, roaming becomes difficult. If the

mobile handset retains its IP address and roams into a different network

(different AS), it would violate the routing constraint that the two different

ASs could not carry the same IP address. If the new operator swaps the IP

address as soon as the handset enters the new network then the TCP/IP sessions

being used by the handset would be naturally dropped. This is an example of some

of the paradoxes faced by juxtaposing Internet and mobility. To overcome some of

these limitations, IETF has drafted a standard called Mobile IP both for IPv4

and IPv6.

Benefits of IPv6

End-to-end addressing

Instead of an almost ubiquitous reliance on Network Address

Translation (NAT) that is required for many networks in the IPv4 Internet,

devices will be able to have their own globally unique address. This will reduce

the amount of end-to-end packet processing and the amount of interference with

upper-layer protocols.

Smaller routing tables

The IPv6 addressing architecture allows for a better

hierarchical design of the Internet. This facilitates a better address

aggregation of the routes, especially at the backbone of the Internet.

Stateless auto-configuration

Stateless auto-configuration makes it possible for

end-stations to configure their own addresses, without the need of a static DHCP

server. Typically, the end-station combines its 48-bit MAC address with a

network prefix it learns from a neighboring router to form its IPv6 address.

Security

The IPv6 packet format includes optional authentication

header and encapsulating security header, providing comprehensive security

features at the network layer. While the format provides authenticity of the

source, the latter provides data integrity and confidentiality.

P Muthukrishnan is

with Juniper Networks