IT Week

IPv6 has been prominent in networking rhetoric and the subject of debate for a number of years but has yet to attain mainstream acceptance. Stuart Mark provides an update on this elusive technology.

IP Version 6

Every industry has its pessimists and networking is no exception. Many apocalyptic predictions have been made since the early nineties regarding the future of IPv4 but here we are in the twenty-first century and the old, reliable protocol is in more demand than ever, thanks to the explosion in Internet usage and the adoption of a single protocol strategy by IT suppliers and customers alike.

Where, then, does this leave IPv6? Hailed as the saviour of networking by some and derided as an unnecessary pipe dream by others, it has been poised on the sidelines since 1995 but has only been officially deployed in the Internet since July last year and is still to find its way into the corporate arena. There are many factors which will regulate the speed of IPv6 implementation but the question of whether that implementation will be total can be answered by looking at the present restrictions of Version 4.

IPv4 Deficiencies

IPv4 was designed for a smaller world and a less technologically active population. It is perceived as a complex protocol but this same complexity has allowed it to evolve to meet the demands of today's Internet-driven culture.

The original protocol specification provides addressing in a 32 bit address space, split into three classes for host assignment and a fourth class for multicasting. Each host class provides a different amount of host addresses and can be chosen according to requirements. This was fine when the Internet was the realm of the nerd and corporate networks used a bewildering mix of proprietary protocols but the swing towards standardisation and the resultant demand for IP addresses has since begun to cause problems.

Subnetting alleviated address shortages as did the specification of Private Address Ranges in RFC 1918 which allowed different organisations to use the same address ranges internally. Network Address Translation (NAT) and firewall technologies resolved most, but not all, inter-company and Internet connectivity issues. IPv4 was also given a reprieve in the Internet by the summarisation advances of Supernetting and Classless Inter-Domain Routing (CIDR) and many pundits still argue that these enhancements negate the requirement for Version 6 or IPng (next generation) as it is sometimes called by providing a hierarchical routing model.

However, there are still many problems associated with IPv4. At 4 billion, the address space is perceived as too finite to accommodate expected growth, especially when so much of the address space has been wasted by inappropriately generous early assignments. Private Addresses and NAT have been used as a get-out-of-jail but these are no more than Band-Aid for anyone serious about Internet connectivity and it's associated security and cost implications.

Summarisation makes routers more efficient by reducing routing tables and introducing a hierarchy but it requires networks to be contiguously grouped. Many address assignments were made before CIDR and are not in any sort of grouping, rendering summarisation unachievable. Also, IP still requires a big investment in administration. Again, this has improved with the advent of Dynamic Host Configuration Protocol (DHCP) and the Domain Name System (DNS) but design is still complex and network changes can still require major DHCP and DNS addressing changes.

Version 6

IPv6 was specifically designed with Version 4 in mind; that is, it tries to resolve as many issues as possible with the incumbent protocol. Of the differences between the versions, the most commonly known is that in address space. IPv6 improves on Four’s 32 bits with a whopping 128 bit address which gives a maximum of 340282366920938463463374607431768211456 addresses! This may seem fantastic but it has allowed designers to provide a much improved addressing hierarchy by dividing the space into aggregators, much like telephone area codes.

At the top of the hierarchy is the Top Level Aggregator (TLA) which generally represents a geographical or Regional Internet Registry area. Below this is the Next Level Aggregator (NLA). These are assigned to large service providers in a geographical context where possible. Three registries provide Version 6 NLAs; ARIN in the Americas, RIPE Network Co-ordination Centre (NCC) in Europe, the Middle East and parts of Africa and APNIC in Asia Pacific.

The large providers can further assign sub-NLA’s to smaller providers if required. Any provider can assign Site Level Aggregators to large organisations who are then free to use the address space as required, safe in the knowledge that it is unique. This method of decentralising address assignment maintains an efficient hierarchy.

At a host level, the address will generally be a concatenation of MAC address (48 or 64 bit) and network ID. This is very similar to Novell’s IPX addressing and carries a clear benefit, autoconfiguration.

IPv6 incorporates a new Neighbour Discovery (ND) protocol that does the job of ICMP and ARP. It is really an extended set of ICMP messages which allows a host to discover its network number to form a host address and its next hop router. It also provides a method of detecting router failures and finding alternate routes. The administration implications of this are clear; no more manual address configuration or DHCP maintenance. Stricter control can be applied, however, by accompanying enhancements to DHCP and DNS. Stateful autoconfiguration is possible using an IPv6 DHCP server and two new DNS records have been added to the specification, A6 and AAAA which provide renumbering improvements, dynamic DNS autoconfiguration and secure DNS.

The version 6 header is a fixed 40 bytes but uses header extensions to provide a number of features. Security enhancements supporting MD5 and SHA-1 are designed to prevent spoofing and packet snooping. Also, the multicast address space has been greatly increased to cope with multimedia applications of the future.

Quality of Service (QoS) is provided for by the inclusion of a ‘Traffic Class’ byte which is a direct replacement of the IPv4 ‘Differentiated Services’ byte. RSVP continues to be supported and there is a new 20bit ‘Traffic Flow ID’ field for enhanced QoS operations.

The protocol makes much more use of multicast. For example ND and DHCP both use multicast to improve communications. In effect, IPv6 does not use broadcasts but there is a new class of address, anycast, which can target a single member of a group. It will be used in redundant configurations.

Implementation

IPv6 is growing on the Internet and has been running on pilot networks such as 6REN and the 6BONE for some years. However, end user implementation is most likely to be vendor-driven. Version 6 stacks are widely available with the big players such as Microsoft, Apple, HP, IBM, Novell and Sun offering developers versions.

Tunnelling, dual stacks and NAT technologies such as RSIP should ease migration but other issues such as application reliance on IP addresses, router filter changes, host file changes which will not be resolved by IPv6 mean that it will not be widely adopted until real business reasons exist. This means the advent of application API’s designed for the protocol.

ISPs will also play a part. They have already begun to use Version 6 in the Internet and it is only a matter of time before the corporate user is targeted. Any resultant translation or tunnelling requirements may make it more cost effective for enterprises to switch to Version 6 for all Internet application. Again, dual v4/v6 stacks may aid in any transition.

There are a growing number of large suppliers which have been assigned NLAs (including BT) and the number of end-system vendors now offering IPv6 stacks makes it clear that acceptance of this protocol is inevitable. If you are a service provider, this won't be news but for corporate users, the coming year is the time to start planning.

Summary

  • IPv6 is designed for a larger Internet environment

  • Security, Performance and Multimedia improvement are built into IPv6

  • Service providers are using the protocol now.

  • In Europe, IPv6 upper level aggregators are assigned by RIPE NCC

  • Corporate networks will receive SLAs from their service provider.

www.ipv6.org

www.ipv6forum.com

www.ripe.net/ripencc/

 

 

 

 

 

       

 

 

 

This site was last updated 04/25/07