Bonus $100
Promo Codes 2024
Users' Choice
90
89
88
85

The IPv6 challenge for telcos and ISPs

26 Jun 2015
00:00
Read More

The day IPv4 addresses run out is happening any day now, with the North American registrar now down to 0.01% worth of /8 address space available. However, thanks to more rigid rationing, the Asia-Pacific region still delayed its exhaustion somewhat. But with plenty of unused address space still left in North America on the commercial transfer market, depletion for now does not mean the end of the world, but rather much higher prices for new projects and an added impetus for moving to IPv6 where possible.

Paul Wilson, director-general of APNIC, the Asia-Pacific Network Information Center, spoke to TelecomAsia of the challenges of transitioning to IPv6 explaining that while telecom executives like to think that it is simply the matter of flicking a switch, behind that switch there are untold numbers of engineers pulling their hair out trying to please their bosses and get everything to work together.

IPv6 transition is non-trivial and potentially expensive.

“If it is as simple as turning it on, why haven’t you turned it on already? Then if not, when actually are you going to flick this mythical switch? If someone says it’s easy their either optimistic or ignorant,” he said.

The ideal model for telcos and ISPs to adopt is a dual-stack approach, that is the operation of both v4 and v6 throughout a given component and ideally that would mean dual-stack use of both address sets within the entire infrastructure.

That would make it the simplest way as the telco would not have to deal with mapping or translation and makes for a more simple DNS configuration.

The only reason that some operators might be opting for something else rather than dual-stack would be that some component of network is v4 or v6 only.

Some phone networks have gone for v6 only and they require address mapping for their v6 handsets. The reason is that some [older] phones’ capabilities are limited and it’s easier for them to run a single protocol stack.

LTE handsets can mostly run dual-stack but older handsets may mean that a single protocol stack is easier.

Another reason is that in 3G, running v4 and v6 together requires a double circuit.

It all boils down to very specific deployment scenarios.

The other group of users struggling with v6 are the CDNs. Wilson said that Google is simply not prepared to subject their users for an extra 0.1 seconds as they measure millions of dollars per millisecond lost. There are so many issues in the way different platforms choose to implement V6 and V4 together.

“There was old Microsoft technology called Teredo, a transition mechanism that was great in theory. But when it was adopted and implemented it simply did not function. The delays involved in handshake and fallback made it useless,” he reminisced.

Even today, every platform has made some strange decision for whatever reason, mostly for reliability or optimising response time, and sometimes this results in very strange behaviour.

“The important thing to remember is that there is no textbook for v6. Even the best practices one year ago can be hopelessly outdated which is why it is so important to focus on our meetings exposing our knowledge and best practices,” he said, inviting anyone in the industry to his events.

APNIC has implemented a rationing system whereby it is still distributing very small amounts of IPv4. Users need to demonstrate a need for v4 addresses for a year before a request is granted. However, Wilson says that ARIN, the American Registry for Internet Numbers,, will run out any day now. What will happen is that they will say no to new requests and put you on the waitlist.

“When that moment comes when ARIN says no, that does not mean a big shift on the Internet. But what it means that ISPs and data centers will be waiting… and you have to choose between IPv6 which is peerable today versus the increased use of NAT.”

ISPs have three choices today - use v4 private addresses with a whole lot of address sharing, deploy expensive carrier grade NAT and end up deploying a substandard solution; go for v4 and pay for the addresses which currently is around $11 an address. The UK government has started selling a fairly large amount that they have had for a long time, but if you want to roll out hundreds of thousands of addresses, that is potentially millions of dollars; or go for the third option is to go pure v6 which has a different set of challenges and costs.

There is a market for v4 transfers, but it is much less of a market than anyone expected. The cost of v4 cannot escalate beyond what’s reasonable; it is not something that someone sensible will invest a lot of cash in.

So will APNIC be transferring address back to ARIN anytime soon? Wilson thinks not. There are still large swathes of underutilised address space in the United States which now has an incentive to be financially liberated. The source of IPv4 addresses will still be from the United States.

Governments as large buyers can help push v6 adoption, but not through the allocation of v6 addresses to individuals to track their online presence. “No, that is not what v6 was designed for,” he said, shaking his head.

.

Related content

Tags:
Rating: 5