By now it is clear to see that the business of networking is not as easy as it might first appear. Before the Internet, organizations simply connected their PCs and other devices together and that was that -- but with the advent of the Internet, suddenly careful attention had to be paid to device addresses and how they were used. Originally, the Internet was envisaged as being for academic institutions, military and government departments. The original addressing plans could cope with those simple requirements but that was before the Internet became the e-business communications backbone that we know today. The problem is compounded by the future where there is a proliferation of mobile devices, all of which will need their own unique IP address, and the time when everything is IP'd, from fridges to car and more beyond.
The growing number of Internet hosts has quickly caused a shortage in IP addresses and will eventually consume the entire address space. The theoretical address space size limitation of 2^32 nodes is largely wasted due to the assignment inefficiencies that we have already covered. Furthermore, dividing the address space into Classes A, B and C contributed to the inefficiency of addresses assignment.
IPv4 is based on binary, so certain arithmetical rules must be maintained. So, you can have 8 or 16 addresses but not 10. Fortunately, the procurement departments of many organizations have yet to realize this and merrily go along buying whatever equipment they need rather than paying any attention to the binary limitations of IPv4. And if they need 10 hosts, organizations will have to maintain a 16-host subnet and lose six addresses. Although IPv4 offers around four billion addresses, there are only 240 million in practical usage.
IPv4 was released in 1980. Twenty years later, in 2000, half of the address space had been used and 74% of the addresses are used by North America. Since 2000, IT has moved quickly into new parts of the world such as India and China. The rest of the world has not stood still either as the number of Internet users increases everyday. For the Internet to work, there is a need for each user or device to have a unique identity -- and so there is a problem with IPv4.
Changing the IP addressing scheme is a matter concerning everyone that accesses the Internet and provides an ideal opportunity for providing additional functionality to the IP so that the next version is more responsive to existing and new demands. The new version of IP is often referred to as IPng; in true Star Trek style this stands for "Internet Protocol: The Next Generation" and will use a hierarchical address structure that should be large enough to meet the needs of the Internet for some time to come.
When it was first realized that something needed to be done there were several proposals for IPng. The most likely to become a standard were -- Common Architecture for the Internet (CATNIP), Simple Internet Protocol Plus (SIPP) and The TCP/UDP over CLNP-Addressed Networks (TUBA) proposals. When reviewing the proposals the main criteria was -- current address assignment policies should not be changed, there is no need to reclaim assigned network numbers and there is no need to renumber most of the Internet. Eventually the Simple Internet Protocol Plus Specification was selected and named IPv6 or IP Version 6.
So, what is in IPv6? The new version of IP provides far larger address spaces for Internet users with the IP address size increased from 32 bits to 128 bits. An address size of 128 bits is large enough to support a huge number of nodes even with the inefficiency of address assignment. This will allow many more nodes than is currently available today, with more levels of addressing hierarchy and simpler auto-configuration of addresses.
The IP header will be changed with some fields being removed in order to keep the overhead low, this is in spite of increasing the number of address bits. Therefore, even though the IPv6 addresses are four times longer than the IPv4 addresses, the IPv6 header is only twice the size of the IPv4 header.
IPv6 will have a new type of address -- a 'Cluster Address', which will identify topological regions. A new function in IPv6 is source routing, which together with Cluster Addressing will allow nodes to control their routing in a more precise way. You can learn more about IPv6 address types and how to transition IPv4 to IPv6 in these two tips that explain how address formats are used to convert IPv4 addresses.
With any change to standards, the big question is just how painful the upgrade or transition will be and the good news is that the transition from IPv4 to IPv6 will be simple and flexible. The upgrade will be incremental with current IPv4 hosts and routers being upgraded to IPv6, whilst new hosts and routers can be installed independently. Backward compatibility is catered for, as existing IPv4 hosts or routers that have been upgraded can continue to use their current IPv4 addresses. The start-up costs are low and minimal effort is needed to upgrade existing systems to IPv6.
What are the benefits of IPv6?
First and foremost, more addresses, lots and lots more. IPv6 uses an address scheme of eight groups of 16 bits to define a 128-bit network address. The address can use hex and uses a colon as a delimiter. If necessary, every device in the world can have a unique address.
IPSec is a set of protocols used for encryption across the Internet and its use is mandatory in IPv6, whereas it was only optional in IPv4. Security in IPv4 data communication existed mainly between two networks -- often only the link between the routers at different locations was encrypted. This changes in IPv6 as its now compulsory to have IPSec at the client level on both sides giving a true end-to-end secure tunnel where security is host-to-host, rather than from network-to-network. (You can learn more about IPv6 network security issues in this podcast.)
Clients using IPv4 addresses will have to use a Dynamic Host Configuration Protocol (DHCP) server every time they log on to a network. IPv6 can allocate permanent addresses that do not need to be resolved by a DHCP server, which creates a plug-and-play environment, simplified management and administration.
To simplify the format, the width of an IPv6 address header is fixed to 40 bits. This compares to an IPv4 header that can be either 20, 40, or 60 bits wide. Quality of Service (QoS) and traffic class services are very limited in IPv4 because there are only three bits available for this. IPv6, in comparison, allows classification and QoS in applications like VoIP. In IPv4, voice got high priority whilst data was a low priority. In order to differentiate all voice traffic on the network IPv6 provides a more granular approach.
IPv6 allows automatic address configuration and reconfiguration meaning that servers can re-number network addresses without accessing all clients. Network Address Translation (NAT) servers may be defunct as there is no need to use private addresses. IPv6 also provides new unicast and multicast methods together with better routing capabilities for mobile devices. There is no broadcast in IPv6.
Of course, one of the first questions for any new piece of technology is how much does it cost? Well, it should not cost anything to move to an IPv6 network. There is no need to buy new hardware and software, all popular operating systems such as Solaris, Red Hat, Unix, Novell, and Windows all have IPv6 stacks built-in. If an earlier version of an operating system is in use, updates should be available free, with the same going for device manufacturers and early routers and switches can be easily IPv6-enabled with a software update.
There is even an opportunity for cost saving as there is no need to dedicate a box to act as a DHCP server and a NAT, meaning those servers can be utilized for other tasks.
When developing a migration plan, organizations should start at the edge. Devices at the edge of an organization's network should run applications that use dual protocol stacks of IPv6 and IPv4. Since many ISPs may not be able to offer IPv6 immediately, the ISP's IPv4 cloud can be used to create a tunnel at an organization's locations to run IPv6 applications. With the edge devices taken care of, its time to slowly move the migration towards the core.
There should be few migration issues although there will be the typical teething problems. The Internet is currently a big cloud of IPv4. As organizations migrate to IPv6, small clouds of IPv6 will appear which will become bigger as the IPv4 cloud shrinks. This may result in some migration or co-existence issues. Work has been done in this area and there are applications that allow co-existence and automatically understand when to use IPv4 and when to use IPv6. These applications implement a dual stack of IPv4 and IPv6 on the same protocol stack so that a host supporting both protocols can communicate with both IPv4 and IPv6 nodes and differentiate between IPv4 and IPv6 packets. Using a dual stack means that existing IPv4 applications will work with IPv6.
As far as network management platforms are concerned, there is currently no support for IPv6. It is expected that network management vendors will provide updates at a later date and any new versions of software will support IPv6.
More on the migration to IPv6
Although, as we have already said, migration should be relatively straightforward, it is worthwhile considering some of the issues that the network administrators will be confronted with as they move their networks over to IPv6.
IPv6 is not backwardly compatible with IPv4 and so an organization moving lock stock and barrel to IPv6 will need to use IPv6-enabled network routers, switches, and hubs, as well as operating systems, applications and other network devices.
This makes the big bang approach to migration financially, as well as perhaps technically, unfeasible for many organizations. The need for such significant investment, in times when the purse strings are being held tight, will mean that the move to IPv6 will be much slower than it would have been 2 or 3 years ago.
The approach that most organizations will take to adopting IPv6 is to running it natively on their backbones. Most will then use IPv6 over IPv4 tunneling for connectivity between the backbone and IPv4 nodes on their own networks, as well as for Internet communications with the outside world. This approach has its drawbacks as tunneling can slow throughput and it also requires network managers to configure information about tunnel endpoints into the encapsulating node, which is a time-consuming process.
Other workarounds can be used; if IPv4 multicast is available, a solution called '6over4' can be implemented where IPv6 multicast is deployed over IPv4 multicast, IPv6 can then utilize neighbor discovery for self-configuration. 'Tunnel Brokering' can be used if IPv4 multicast is not available and uses dedicated servers to configure tunnels on behalf of remote IPv6 clients. With another play on the numbers, '6to4' allows single hosts or small domains on IPv4 networks to communicate with IPv6 nodes with minimal need for manual configuration. The service providers are also getting ready for the change as NSPs (network service providers) are running IPv6 with limited deployments.
There is more work to be done in the area of the applications that will need to be upgraded to take advantage of IPv6's features in Quality of Service and mobile IP addressing. Software will also need to be upgraded to take advantage of IPv6's new security features as well as the ability to communicate with IPv6 hosts.
Whichever approach is taken it all comes down to planning, and as everyone knows the more time spent up front in any project the more chance of its success.
View our IPv6 tutorial for more information, or read this Q&A on who needs IPv6: Does your business network need an IPv6 transition?
Copyright 2002. IT-Director.com provides IT decision makers with free daily e-mails containing news analysis, member-only discussion forums, free research, technology spotlights and free on-line consultancy.