June 4, 2013 – Anyone taking comfort in the assumption that delaying the transition to IPv6 won’t incur considerable cost penalties should get in touch with Lee Howard, director of network technology at Time Warner Cable.
Howard, who has taken a deep dive into the options and costs of procrastination, suggests the costs could be astronomical, depending on how long the inevitable jump to IPv6 is put off. While acknowledging “there’s a big spread in projections” as to when IPv4 addresses will no longer be available from the pool allocated by the American Registry of Internet Numbers (ARIN) and other regional Internet registries (RIRs), Howard makes clear the costs go up with every passing day that preparations for the transition aren’t in sway no matter when the v4 addresses run out.
His analysis comes as a response to widespread speculation that it will be possible to buy from large stocks of unused or under-utilized IPv4 addresses. Costs to be incurred on this track are anybody’s guess, but they’re sure to go up as the supply goes down, he says.
The bottom line is that, given the inevitability of eventual conversion to IPv6, whatever the costs of stocking up on second-hand IPv4 addresses turn out to be, they’ll be additive to the conversion costs. “If you want to know how to reduce costs, it’s pretty obvious,” he says. “Deploy v6 sooner than later.”
According to Howard’s research, there remain about 144 million v4 addresses that have yet to be allocated by the five RIRs, which will likely run out by the middle of next year. The number of unused v4 addresses worldwide comes to about 480 million, while there are another 520 million underutilized addresses, which is to say, addresses that have been handed out to users who are no longer on the distributor’s network.
Howard goes so far as to whimsically include another 1.5 billion addresses, comprising the rest of the roughly 2.7 billion v4 addresses in the allocated public address pool (which doesn’t include roughly 1.5 billion held in reserve for governments and other purposes) as “substitutable” through replacement by IPv6 addresses and/or use of the Network Address Translation (NAT) protocol, which assigns private addresses to devices running behind publicly addressed devices. In other words, for the sake of argument, he suggests nearly the entirety of the allocated v4 address pool could theoretically be resold to IPv6 laggards.
As to what the costs might be for addresses falling within each of these categories, Howard bases his estimates for all but the substitutable category on existing market rates. Costs of the as-yet-unallocated v4 addresses, which are free but incur costs for management and storage, range from 3 cents to $4 per address. The going rate for sale of unused addresses is $9-$12 per address, while the underutilized addresses, counting costs of revamping the sellers’ numbering set-up, are in the $10-$16 range.
Whether addresses in the last category might actually be sold depends on how forceful a company might be in persuading its CIO to undertake the renumbering hassles involved, Howard says. “You might have to tell the CIO he’s going to be driving a Bentley if he does this,” he jokes.
As for prying loose the “substitutable” numbers already in regular use, maybe $100 or more per address would do the trick, he suggests. Estimating what the demand curve might look like as ever more v4 addresses are needed to serve the expanding user and device population against a dwindling supply, Howard puts the average cost, not counting the “substitutable” category, at $9-$12 in 2014, $9-$16 in 2015 and $16-$20 in 2016, with the supply in all but the substitutable category running out by 2017.
At that point everyone will have to begin the transition to IPv6, which will start incurring either the costs of Carrier Grade NAT (CGN) for those who are at risk of running out of IPv4 addresses before they’ve made the transition to all-IPv6 or the costs of dual stack operations, where IPv6 is used to connect new subscriber and there’s enough of a v4 reserve to cover the address requirements of v4-only devices that those new subscribers bring onto the market.
At some point, because all new devices are running IPv6, those older devices will fade out of the picture, allowing operators to use IPv6 exclusively for all new subscriber addressing requirements. By moving immediately to end-to-end dual stack operations allowing use of IPv6 for all new connections wherever possible, operators will avoid depleting the v4 reserves they need for those new subscribers, thereby avoiding the need to implement CGN.
This is vital to keeping costs at a minimum. Howard calculates the CAPEX and OPEX costs for using CGN in conjunction with IPv6 will come to $30 per new subscriber per year in contrast to about $7.50 per new subscriber per year for the dual stack approach. For those who decide to keep buying v4 addresses on the resale market until they run out, the $9-$20 per user costs of those addresses will have to be added to the eventual costs incurred with either CGN or dual stack.