220 likes | 238 Views
Explore IPv4 allocation and usage trends, with a focus on AS number exhaustion and forecasting solutions. Learn about exponential vs. quadratic growth patterns and the need for proactive forecasting to address challenges like route table bloat. Discover projections, growth rates, and lessons for policymakers in the realm of IP address allocation.
E N D
IPv4 Address Allocation Trends J. Scott Marcus Chief Technology Officer (CTO) May 22, 2001
IPv4 Address Allocation Trends • AS number exhaustion - the RIRs recognize the need for forecasting • IPv4 allocation and usage trends • The McFadden/Holmes Report
Exponential Growth of Autonomous System (AS) numbers Source: Scott Marcus, Genuity
Chicken Little was Wrong! • This is far simpler to remedy than IPv4 address exhaustion, because • the solution need not impact end systems (hosts); • the solution need not impact DNS; and • the solution need not impact routers unless they speak BGP-4. • Any solution is complicated by the need for backward compatibility and phased migration. • Time until exhaustion is nonetheless sufficient to architect, design, implement and deploy solutions. • Cisco and Juniper are reportedly well into implementation.
Route Table Bloat - a Different Problem BGP Table Growth since 1989
The RIRs Recognize the Need for Forecasting • Continuing need to further refine projections. • Need for forward-looking proactive forecasting on a regular basis not only for AS numbers, but also for route table entries and IPv4/IPv6 addresses. • Forecasting needs to incorporate allocation data from all three RIRs (APNIC, ARIN, RIPE NCC). • Forecasting needs to be institutionalized by the RIRs themselves, with data readily available to independent researchers.
The Team • Assembled by ARIN • Frank Solensky Gotham Networks • kc claffy CAIDA • Scott Marcus Genuity • Active contributions and support by APNIC and RIPE NCC
Goals (and Non-Goals) for RIR Team • Formally and regularly: • Gather address deployment data from RIRs • Perform statistical analysis • Make allocation data available for independent analysis • Non-goals: • ‘Blue sky’ theorizing • Estimating NAT utilization (for now)
McFadden/Holmes/Mylotte Projection Purported Worst Case, Most Likely and Best Cases
Rate of Depletion • Annual growth rate • McFadden/Holmes assumes low/medium/high growth of 30%/50%/80% respectively. • RIR data shows annual growth of about 3%, with a negative second derivative • Geoff Huston’s data shows annual growth of roughly 7% • Order of magnitude discrepancy! • The trend does NOT appear to be exponential overall.
Marketing Analysis • Numerous statistics, no attribution • No traceability of assumptions • No verifiability of source data
Lessons • The RIR team will continue to focus on conservative analysis and extrapolation of verifiable quantitative data. • Debate is a healthy thing - “Let a thousand flowers bloom.” • We will make RIR source data available to independent researchers. • Policymakers in the RIRs, ASO, IETF and elsewhere should benefit from diverse inputs and should reach their own conclusions.
Acknowledgments • Frank Solensky, Gotham Networks • kc claffy, CAIDA • Geoff Huston, Telstra • Cathy Murphy, ARIN • Paul Wilson, APNIC • Axel Pawlik and Mirjam Kuehne, RIPE NCC • Tony Bates and Phillip Smith, Cisco • Mark Kosters, Verisign • Christian Huitema, Microsoft