Community meets to discuss the challenges
By: Henri Wohlfarth
Date: October 7, 2007
At a joint IESG-IAB meeting, participants discussed the deployment of IPv6, the state of the IPv4 address pool, the challenges of both, and what the IETF can do about it all.
While predictions about timing may vary, there is virtually no disagreement that IPv4 addresses will, at some point, become unavailable. This concern has been the subject of debate and discussion among participants of the IETF and others for more than two decades. It also has been the key driver in the development and deployment of a number of new technologies, including IPv6. The IETF Journal was invited to join the discussion as part of a joint IAB-IESG meeting held in July in Chicago at IETF 69. What follows is a report by the IETF Journal on what transpired during that single meeting, one of several on this important topic. The discussion was led by Kurtis Lindqvist and Jari Arkko.
IPv4 Address Space Allocation and Policies
IPv4 PA (provider-aggregatable) allocations flow from IANA to the RIRs (Regional Internet Registries), which include the RIPE NCC, ARIN, APNIC, LACNIC, and AFRINIC. From there, address space flows to the LIRs (Local Internet Registries) and then trickles down to the end users.
Similarly, IPv4 PI (provider-independent) allocations begin with IANA and then flow to the RIRs. From there, address space is assigned directly to end users.
Each RIR is responsible for forming its own address allocation policy within its regional community in an open, bottom-up process. Today, IANA allocates address blocks of /8 to the RIRs, and it plans to continue allocating sufficient IPv4 address space to support RIR registration needs for at least 18 months. Visit the IANA allocation policy (PDF).
When Will the IPv4 Address Space Become Exhausted?
In May 2007, Geoff Huston announced that his model for determining the projected date of IANA unallocated IPv4 address exhaustion has changed. As reported at www.potaroo.net, the new predictive model is based on a quadratic equation, which, he believes, offers a closer fit to the underlying data set. Here the exhaustion point is the date that the first RIR exhausts its available pool of addresses and no further addresses are available in the IANA unallocated pool to replenish the RIRs’ pool. The data available suggests a best-fit predictive model whereby this will occur in January 2011. A related prediction is the exhaustion of the IANA unallocated number pool, which this model predicts will occur in June 2010.
In a report titled A Pragmatic Report on IPv4 Address Space Consumption, Tony Hain describes a slightly different methodology for arriving at predictions regarding the time constraints for the exhaustion of IPv4 address space. As Tony writes, “Depending on the model chosen, the nonlinear historical trends . . . covering the last 5- and 10-year data show that the remaining 64 /8s will be allocated somewhere between 2009 and 2016, [assuming there is] no change in policy or demand.”
Regardless of the differing methodologies and assumptions, both Geoff and Tony have been vocal advocates for the need to “commence investment in IPv6-based service infrastructure,” as Geoff describes it in a response to Tony’s projections.
The abounding sense of urgency related to the deployment of IPv6 within the IETF community, as Kurtis points out, affects-and is affected by-both policy and market forces. “The dates tell us how much time we have, and the RIRs must come up with a good model or a process to handle [the transition],” he says. As one attendee mentioned, RIRs are allocating addresses under the current model, and as long as the model doesn’t change dramatically, reasonable predictions can be made, but the RIRs are beginning to see policy change requests, which could alter those predictions.
The question that is most on the minds of the IETF community is, What will it take for the market to begin the deployment of IPv6? At some point, the costs of obtaining IPv4 address space will be higher than the costs of deploying IPv6. Transition to IPv6 is fraught with challenges-including the expense of new equipment- but the biggest challenge appears to be that there is no seamless way to transition back and forth from IPv4 to IPv6.
IANA and the RIRs are actively encouraging their communities to deploy IPv6. ARIN and LACNIC, as well as ICANN, have all published statements promoting the move to IPv6. While the proposal by ARIN to set a date for IPv4 termination was not adopted, the discussion within the RIRs remains active.
Implications for the Future of the Internet
As Kurtis and Jari explained, perhaps the most significant implication of IPv4 exhaustion on the Internet is that NAT (network address translation) is “here to stay.” Other predictions regarding the implications of IPv4 address- space exhaustion for the next phase of the Internet include:
- Some IPv6 deployment
- Some kind of market space forming for addresses
- Security of routing becoming more interesting
More routing pain; smaller prefix blocks Some of the political implications are likely to include:
- The appearance of last-chance-allocation panic
- Discussions about address allocation policies
- Debates about fairness between different parties, such as old/new users, different registries, IANA versus registries, and developed world versus developing world
- Market creation to be driven by political and legal battles
What Is the Role of the IETF?
Deciphering the role of the IETF with regard to IPv4 address space depletion and the deployment of IPv6 was a large part of the discussion at the IAB-IESG meeting. Kurtis and Jari outlined a number of aspects that fall outside the IETF’s purview, such as consideration of the allocation rate of existing address space (a discussion that should happen within the RIR communities) and policies that affect how that address space gets allocated (also the responsibility of IANA and the RIRs, which all have active and open discussion forums for debating those policies). Other aspects not applicable to the IETF include the possible address space markets (which should be left to registries, contracts, and the courts) and NATs and IPv6, which may well be IETF issues.
The area that Kurtis and Jari said they believe does fall under the auspices of the IETF is delivery of the technical components that the address space market will require. As they pointed out, NATs are “a fact of life,” and they should continue to be taken into account. The deployment of any significant new technology on the Internet, such as IPv6, “is going to be painful.” Both suggested it may be time to take another look at IPv6 transition mechanisms. “Things related to IPv6 will have to be fixed as deployment goes on,” said Jari. “This is maintenance work, and the IETF is good at that.”
Kurtis agreed, adding that transition mechanisms exist but they often require purchasing new equipment, handling implementation issues, and training staff, all of which requires considerable investment. What the IETF may need to consider is whether there is anything it can do to make the transition easier. “This is what we have to solve,” Kurtis said.
According to Joe Abley, the perception is that the only generic transition mechanism is dual stack, but the dual stack solution could double the cost for companies and organisations. “In many cases,” he said, “it is easier to implement only IPv6, but there is no good way to get back to the IPv4 Internet from there.” Achieving that, he said, could solve the problem.
Elwyn Davies suggested that in light of there being no interworking transition mechanism, it might make sense to go back and develop one that works better than NAT-PT (network address translation-protocol translation). “It would be easier if IPv4 were a special case in IPv6,” said Elwyn. “Then we would not have the interworking issues between the two networks. But that is not the case.”
So what can the IETF do about the diminishing IPv4 address space in and the smoothing of the transition to IPv6? Distributing the remaining IPv4 address space is the responsibility of the RIR community. If the RIRs need any technology support, such as in the case of classless interdomain routing (CIDR), the IETF is there to help. With regard to the transition mechanisms, it would probably be best to wait and see what the market is doing. If it turns out that the existing mechanisms don’t work or that people have difficulties with them, then the IETF will have to look at that.
It may be true that there is little the IETF can do apart from encouraging the market to be proactive in testing and deploying IPv6. However, what drives the market today are economics and regulation. With that in mind, Dave Thaler suggested that the IETF consider providing review of the technical setup and requirements of government agencies. “IETF participants or working groups could help find out what the market demand is going to be-particularly by looking at what large government agencies are rolling out,” he said. “A lot of the other things are outside the scope of the IETF.”