IPv4 address exhaustion — the actual running-out of unallocated IPv4 addresses — happened at the IANA level in February 2011 and at the regional registry level progressively through 2019, and the mechanisms that have kept the internet running since (NAT, IPv4 trading markets, and carrier-grade NAT) each introduce specific networking problems that IPv6 was designed to eliminate
Understanding subnetting is relatively straightforward — CIDR notation, host counts, network masks. What's less discussed is the broader context: why subnetting exists at the scale it does, what the IPv4 address shortage means for network architecture decisions made today, and how IPv6's enormous address space changes the economics and architecture of internet connectivity.
How IPv4 exhaustion actually unfolded
The hierarchical allocation structure:
- IANA (Internet Assigned Numbers Authority): the top-level body that allocates /8 blocks (16 million addresses each) to the five Regional Internet Registries
- RIRs (Regional Internet Registries): ARIN (Americas), RIPE NCC (Europe/Middle East/Central Asia), APNIC (Asia-Pacific), LACNIC (Latin America), AFRINIC (Africa) — allocate to ISPs and organisations
- ISPs and organisations: allocate to end customers and networks
The exhaustion timeline:
- February 3, 2011: IANA allocated its last five remaining /8 blocks (one to each RIR) and declared the free pool exhausted
- April 2011: APNIC exhausted its allocation, entering "final /8 policy" (each new applicant gets at most a /22, approximately 1,000 addresses)
- September 2012: RIPE NCC exhausted its allocation
- June 2015: ARIN exhausted its allocation
- August 2020: LACNIC exhausted its allocation
- AFRINIC: entered its final /8 policy in 2021
What "exhausted" means in practice: new organisations can no longer receive a block of IPv4 addresses from their regional registry. They must either purchase addresses from other organisations on the IPv4 trading market (prices peaked around $50-60 per address in 2021-2022) or use IPv6.
Carrier-Grade NAT: the stopgap that breaks internet assumptions
Standard NAT (as used in home routers) maps one public IP address to many private devices. Carrier-Grade NAT (CGN, also called Large-Scale NAT or LSN) adds another layer: the ISP places many customers behind a single public IP address, translating between the customer's private address space and the ISP's shared public IP.
The topology: Customer device → Customer router NAT (private → private) → ISP CGN (private → public) → Internet. Two full NAT translations for every connection.
What CGN breaks:
Peer-to-peer connectivity: applications that rely on direct connections between two endpoints (VoIP, gaming, BitTorrent, WebRTC) fail or degrade significantly through double NAT because neither endpoint has a publicly reachable IP address.
Port forwarding: impossible through CGN. Customers cannot host services or use port-forward-dependent applications (some VPN configurations, remote desktop, home servers).
IP-based geolocation accuracy: CGN concentrates thousands of customers behind a single IP, making IP-based geolocation and fraud detection significantly less accurate — all those customers appear to be at the ISP's CGN device location.
Regulatory compliance: logging requirements (which IP address made which connection at which time) require ISPs to log the customer source IP and port plus the public IP and port — significantly more complex logging than traditional NAT, and a compliance challenge for law enforcement.
IPv6 subnet scale: why /64 is the smallest allocatable unit
IPv6 has 128-bit addresses compared to IPv4's 32-bit — producing 2^128 ≈ 3.4 × 10^38 total addresses. To contextualise: if you divided all IPv6 addresses among every person on Earth, each person would receive approximately 4.8 × 10^28 addresses.
The IPv6 subnet size conventions:
/32— typically allocated to ISPs by their RIR/48— typically allocated to individual organisations or large sites/56— sometimes allocated to residential customers (gives 256 /64 subnets)/64— the standard subnet size for a single network segment, regardless of how many devices are on it
Why /64 is the minimum for a connected segment: IPv6's SLAAC (Stateless Address Autoconfiguration) — the mechanism by which devices automatically generate their own IPv6 address without DHCP — uses the /64 prefix combined with a 64-bit interface identifier (historically the device's MAC address via EUI-64, now typically privacy-randomised). SLAAC only works with /64 prefixes; subnetting more finely than /64 breaks SLAAC.
The abundance implication: a home customer with a /56 gets 256 /64 subnets — more than enough for separate subnets for IoT devices, computers, guest network, home lab, and any future use case, with every subnet having 2^64 possible addresses. The concept of "conserving" addresses within a subnet is obsolete under IPv6.
The transition mechanisms: how IPv4 and IPv6 coexist
Dual-stack: a network device has both an IPv4 and an IPv6 address and can communicate on either protocol. A dual-stack web server serves IPv6 clients over IPv6 and IPv4 clients over IPv4. This is the cleanest transition mechanism but requires IPv4 addresses to remain available.
6to4 and Teredo (largely deprecated): tunnelling mechanisms that carried IPv6 packets inside IPv4 packets, allowing IPv6 connectivity without native IPv6 support from the ISP. These introduced latency and complexity and are increasingly deprecated.
NAT64/DNS64: allows IPv6-only clients to reach IPv4-only servers. The DNS resolver returns a synthetic IPv6 address embedding the IPv4 address; the NAT64 gateway translates between IPv6 and IPv4. This enables full IPv6-only networks (eliminating IPv4 entirely, including CGN) while maintaining access to the IPv4 internet.
Subnetting decisions in a dual-stack world
Modern network design for new deployments should plan for IPv6 natively rather than treating it as an add-on:
For cloud VPCs: both AWS and GCP support dual-stack VPCs where each subnet has both an IPv4 CIDR (e.g., 10.0.1.0/24) and an IPv6 CIDR (e.g., 2600:1f18::/64). Resources can be assigned IPv6 addresses in addition to or instead of IPv4.
For home and office networks: request a /48 or /56 from your ISP (many provide this, especially if you ask) and assign /64 subnets to each network segment — allowing future expansion without renumbering.
For IoT: the argument for IPv6 is particularly strong here — IoT devices benefit from end-to-end addressability (eliminating NAT traversal complexity for device management) and the enormous address space avoids any constraint on the number of connected devices.
How to use the Subnet Calculator on sadiqbd.com
- For IPv6 planning: the calculator handles both IPv4 and IPv6 CIDR notation — enter a /48 prefix to see the number of /64 subnets available within it, or enter a /64 to understand the host address range (which is effectively unlimited for practical purposes)
- For CGN impact assessment: if your ISP uses CGN, you're sharing a public IP — verify your public IP by checking the calculator's output for your gateway address and comparing against whatismyip services; if your router's WAN IP is in the 100.64.0.0/10 range (IANA's designated CGN space), you're behind carrier-grade NAT
- For dual-stack subnet design: plan parallel IPv4 and IPv6 subnet structures where the IPv6 /64 corresponds to each IPv4 /24 or /28 — maintaining conceptual clarity between the two addressing schemes for the same network segments
Frequently Asked Questions
If IPv4 addresses are "exhausted," why can I still get an IPv4 address from my ISP? Because ISPs stockpiled large allocations before exhaustion, and the trading market allows transfers. ISPs that obtained large IPv4 allocations in the 1990s and 2000s (when the address space was being given away rather than sold) have reserves that are being gradually depleted as customer demand grows. New ISPs or those with insufficient stockpiles must purchase IPv4 addresses on the secondary market at the current market price. Eventually, continued IPv4 address availability depends on the trading market transferring addresses from under-utilising legacy holders (some universities received enormous early allocations like /8s they don't fully use) to ISPs with active demand.
Is the Subnet Calculator free? Yes — completely free, no sign-up required.
Try the Subnet Calculator free at sadiqbd.com — calculate network addresses, host ranges, and broadcast addresses for any IPv4 or IPv6 subnet.
