IPv6 is now 15 years old. By observable measures (scientific or anecdotal), the successor to IPv4 still appears to be crawling on hands and knees. A recent study by Dhamdhere, Luckie, Huffaker, Claffy, Elmokashfi and Aben, however, offers evidence that tiny toddler IPv6 appears ready to walk, or at least cruise.
Measuring the Deployment of IPv6: Topology, Routing and Performance uses historical data sets of BGP traffic collected by the University of Oregon (RouteViews Project) and RIPE (Routing Information Service) to analyze the growth and performance of predominantly used customer-provider links of the IPv6 network. These data sets are extensive and the methodology is well-documented, conceived, and executed.
The authors' findings that I found most interesting include:
- Enterprise adoption is growing. The fraction of networks representing enterprise customers - one of four business types the authors choose to characterize and trend growth - is increasing steadily ("In 2003 only 35% of IPv6 networks were of type EC, but this fraction has increased steadily, currently at 60%."). I suspect that availability of IPv6 capable middleboxes, exhaustion of IPv4 allocations, and availability of IPv6 enterprise solutions from top-tier ISPs are important enablers here.
- Growth in edge networks is lagging behind the core, but now catching up. The article calls attention to the growth in enterprise customers. I suspect that persistent sluggish growth in edge networks will have much to do with the lack of access provider economic incentives, availability of consumer grade (and commodity priced) IPv6 routers, or the temptation to do carrier grade NAT and forestall IPv6 deployment for as long as possible.
- North America is stepping up to IPv6. Europe and Asia were earlier adopters of IPv6, partly because IPv4 address exhaustion hit harder there. Howeer, the authors found that, "While the RIPE and APNIC regions led early adoption of IPv6, adoption in the ARIN region is accelerating, and the number of ARIN-registered and RIPE-registered ASs in IPv6 currently grow at the same rates". More importantly, the authors found that, "IPv4 now grows linearly in terms of ASs and prefixes, while IPv6 grows exponentially." I suspect that reducing the economic and technology inhibitors for access providers (and in so doing, convincing them that carrier grade NAT is a questionable investment) will be essential to sustain exponential growth over a long horizon.
- Routing paths are a critical factor in IPv6 performance. The authors found that IPv6 performance is comparable to IPv4 where routing (AS-level paths) are the same, but much worse when the IPv6 paths are different. More importantly, they found that "while only 40-50% of AS paths are currently identical in IPv4 and IPv6, up to 95% of AS paths could be identical, if current IPv6-capable ASes established equivalent peerings in IPv4 and IPv6." This is an signal find and providers should take note. (Please also see Assessing IPv6 Through Web Access - A Measurement Study and Its Findings.)
Comments