Why do we need a new internet protocol?

If it’s not broke, why fix it? Simple... For the internet to function as designed, every host needs a globally unique address.  IPv4 (the current protocol) supports a bit less than 4 billion such addresses.  When the expectation was that a few research institutions, universities, military installations, etc. were going to need IPv4 addresses and little else, this was a vast, seemingly infinite space. However, the World Wide Web changed all that in the mid-1990s and now it doesn’t take a lot of math to realize that with 6 billion+ people, cellphones, laptops, desktops, servers, web sites, etc. all needing addresses, IPv4 runs out fairly soon.

To fully understand the nuances of runout is beyond the scope of this paper, but, some basics are in order here.  IP address uniqueness is maintained by a central assignment authority called IANA (Internet Assigned Numbers Authority). IANA hands addresses out to a group of Regional Internet Registries (RIR) in very large blocks (about 16.7 million addresses per block) called “slash-8s”, usually written as /8. There are currently 5 RIRs.  When IANA has only 5 /8s remaining, IANA will give one /8 to each RIR.  This event is expected to occur some time in late 2011 based on current projections.

Once IANA runs out, the RIRs will probably be able to continue making allocations for 1-2 years (the more active regions such as Europe, Asia-Pacific, and North America running out sooner).

Shortly after RIRs run out of space, Internet Service Providers (ISPs) will start running out.  When this happens, new customers connecting to ISPs will have to be more creative than merely asking for addresses. Possibly purchasing them on a transfer market, possibly using only IPv6, etc.

The author believes that the first IPv6-only networks will be in the residential broad-band market (cable, FIOS/BPON/GPON, DSL, etc.). While the service providers will provide IPv6 to IPv4 gateway services, these users will experience various levels of service impairment when connecting to IPv4-based services.

As such, services (web sites, instant messaging systems, VOIP providers, etc.) which are ready for IPv6 before this happens will be in a much better position to serve these customers than competitors who choose not to implement IPv6 earlier.

Considerations for Change

There are those that will argue that there is no business case for going to IPv6 before you have to. Others believe that the sooner you start implementing IPv6, the more advantage can be gained over the competition.

Either way, there are many things that must be considered in this process. First, there are the technology issues. All your routers, switches, load balancers, servers, etc. need to be IPv6-ready. In some cases, this will require hardware upgrades. In a few cases, software may need to be updated. In most cases, existing infrastructure will be mostly, if not completely, IPv6 ready.

Once the technology is ready, your deployment staff needs to get up to speed. This will include training, research, planning. etc. Then, IPv6 can be deployed alongside IPv4 on existing networks, first in small test areas, gradually expanding to your entire network infrastructure. Once deployed to a sufficient portion of the environment, adding IPv6 capabilities to servers and the services they provide can begin. This is also a likely time to start looking at internal software which requires updates.

Finally, training of support staff and other personnel.

The key thing to remember here is that adding IPv6 services to the network can, and should be accomplished in a way which does not interfere with existing IPv4 capabilities. This is not (at least initially) a conversion from IPv4 to IPv6. It is an change from IPv4-only to IPv4/IPv6 dual-capability (known as dual-stack).

Next: How to Change

Days