Talk to a specialist

BLOG VLSM.io

IPv4 exhaustion

Understand why IPv4 has become scarce, how its limitations affect applications, and how to plan operational continuity and growth with VLSM.io solutions.

IPv4-only egressConnectivity and intelligence
Globe with IPv4-only outbound connections
IPv4-only egress203.0.113.24Illustrative addresses
Estimated reading time: 10 min9 sections
IPv4 exhaustion
IPv4 exhaustion
01

What ran out: the address pool, not IPv4 connectivity

IPv4 identifies network interfaces using 32-bit addresses. This allows 2³² combinations, or 4,294,967,296 addresses in the theoretical space. That number is not a pool of available public addresses: parts of the space serve specific purposes, including private networks, loopback, multicast and documentation. IANA maintains records of these reservations. Internet growth has created demand beyond what this model can individually provide to all new connections, businesses and applications.

Exhaustion means that pools for ordinary allocation have reached their limits or become subject to restrictive policies. Addresses already in use continue to work; there is no worldwide IPv4 shutdown date. The constraint becomes most visible when additional resources are needed. More efficient allocation with CIDR or address sharing through NAT extends use of the existing space without enlarging the protocol’s 32-bit address field. IPv4 availability therefore belongs in expansion planning alongside processing capacity, connectivity and service availability.

CONNECTIONS IN MOTIONThe path of your connection
ApplicationJob or integration
ConnectivityIPv4-only egress
DestinationSystem or service
Example of a service with IPv4-only Internet egress. IPv6 egress belongs to a separate service. The animation does not represent live traffic.
02

Official IPv4 exhaustion milestones around the world

IANA coordinates the global pool, while the five Regional Internet Registries, or RIRs, administer resources in their regions. APNIC, ARIN, AFRINIC, LACNIC and the RIPE NCC reached different stages on different dates. Reaching a final reserved block, restricting allocation and exhausting an available pool are different events. The milestones below describe those historical events, rather than a real-time measurement of addresses that each registry may still recover or distribute.

At APNIC, reaching the final /8 in 2011 triggered conservation rules, rather than unrestricted supply until a shutdown date. AFRINIC’s January 2020 milestone marks phase 2 of restricted allocation, known as soft landing; it must not be described as a completely empty pool. ARIN, the RIPE NCC and LACNIC also maintain post-exhaustion mechanisms under their respective policies. Comparing regions using only the word “exhausted” hides differences that matter when designing a network or procuring connectivity.

DateRegistryMilestone and meaning
3 Feb 2011IANAFinal five /8 blocks distributed to the RIRs; central free pool exhausted.
15 Apr 2011APNICFinal /8 reached, with allocation limited by regional policies.
24 Sep 2015ARINFree pool exhausted; requests depend on policy-defined mechanisms.
1 Mar 2019IANAFinal recovered-pool allocations under the global mechanism: two /24s to each RIR.
25 Nov 2019RIPE NCCRemaining pool exhausted; recovered resources serve the waiting list.
13 Jan 2020AFRINICSoft-landing phase 2 begins: restricted allocation, not an announcement of an empty pool.
19 Aug 2020LACNICRegional pool exhausted; limited continuation through returns, recoveries and specific reserves.
03

Brazil: what the waiting list can and cannot provide

Brazil is part of the LACNIC region, with number-resource services provided through NIC.br/Registro.br. The regional pool was exhausted on 19 August 2020. Official guidance explains that returned or recovered blocks may become available after quarantine. However, neither recovery timing nor quantity can be predicted. A project with a fixed launch date should therefore not treat a waiting-list request as immediate delivery or as capacity already available for production.

The guidance consulted describes eligibility criteria, justified need, and ASN and IPv6 allocations as part of entering the Brazilian waiting list. Approval and waiting are administrative stages distinct from activating a network.

Existing addresses may continue serving applications through their current holders or arrangements allowed by policy. Businesses should distinguish contracted and delivered capacity from resources dependent on review, future availability or a provider change. This prevents commitments to new customers from relying on resources that are still uncertain.

TRAFFIC AND ADDRESSINGObserve the source and track responses.
Source203.0.113.24
Destination198.51.100.20
Return203.0.113.24
Addresses reserved for examples. Track bidirectional traffic together with latency, failures, and application results.
04

NAT and CGNAT: continuity with technical limits

NAT translates addresses and, in implementations that share ports, allows multiple devices to use one public IPv4 address. CGNAT applies that sharing at provider scale. APNIC explains that this extends IPv4 usability, but IPv4-to-IPv4 NAT is not itself a transition to IPv6. It manages scarcity within the existing address space while introducing equipment and rules that become part of the connection path.

This design requires capacity planning for ports, concurrent connections, processing and translation state. More users may compete for resources in the same mechanism. Inbound connections and applications that depend on end-to-end communication can require specific configuration or encounter limitations. Port forwarding on the customer’s router alone does not control the provider’s additional translation layer. The impact depends on architecture and application behavior: neither universal failure behind CGNAT nor complete transparency in every scenario is an accurate assumption.

For automation, a practical consequence is the need to distinguish a public address from an exclusive origin. Several operations may appear to a destination under the same IP, and troubleshooting must consider timing, ports and session context. Dedicated egress can help when a project requires its own origin, but it does not replace application testing. For RPA and data collection, evaluate authentication, queries, downloads and session duration as well as basic connectivity.

05

Allocation, transfer and service procurement

Receiving a block from a registry, completing a recognized transfer and contracting use of a provider’s addresses are different paths. ARIN maintains guidance on post-exhaustion requests and transfers; other registries apply their own rules. A transfer redistributes existing resources rather than creating new IPv4 addresses. Similarly, renting a subnet or using a proxy is not automatically equivalent to receiving an allocation directly from an RIR. Identify the actual resource or service being offered.

When assessing a service, confirm prefix identification where applicable, authorization to use it, delivery method, capacity, support and replacement or termination procedures. A proxy serves applications configured to use that intermediary; a prefix delivered to infrastructure involves its own routing requirements. Addresses, transit and bandwidth are not interchangeable. The RIPE NCC explains that transfers and sharing can help in the short term without solving the structural shortage of IPv4 space. Procurement decisions should account for that constraint.

06

IPv6 expands the address space but requires destination support

IPv6 uses 128-bit addresses, greatly expanding the space available for growth. Adoption should be planned around actual applications and dependencies. IPv4 and IPv6 do not communicate directly merely because their names are similar: coexistence, translation and proxy intermediation are different mechanisms. IPv6 support on the source network does not automatically make an IPv4-only system reachable. Checking a portal’s homepage is also insufficient when authentication, APIs or files use other destinations.

For automation, record which endpoints accept each family and test complete workflows. Retain IPv4 where compatibility still requires it and use IPv6 for workflows with verified support. This reduces dependence on the scarce resource while preserving essential integrations. With a proxy, inspect the application-to-proxy connection separately from the proxy-to-destination connection. Ingress and egress may use different address families; the service’s egress determines compatibility with the destination.

07

How VLSM.io supports continuity and growth

VLSM.io’s approach combines addressing requirements with the way your operation works. For RPA, integrations and data collection using a proxy, we offer IP Proxy through IPv4 Proxy and IPv6 Proxy services. Both accept application connections through IPv4:Port. IPv4 Proxy reaches the destination over IPv4, while IPv6 Proxy uses IPv6 egress. An application able to reach the IPv4 ingress can therefore use IPv6 Proxy for a compatible destination without native IPv6 connectivity on that first leg.

Both families offer static, rotating and random modes. Static mode keeps the egress IP fixed; rotating mode changes it at the configured interval in minutes; random mode uses an egress IP different from the preceding request for each new request. The ingress IPv4 and port remain configured in the application.

Match the mode to the continuity the process requires: an integration using an origin allowlist may need static egress, while other workflows can support planned egress changes. Rotation redistributes use of service addresses; it does not create IPv4 addresses or enlarge the global address space.

For projects requiring address blocks, VLSM.io also offers IPv4 and IPv6 subnet rental with delivery sized for the infrastructure. The proposal should define availability, block size, bandwidth, integration and support. The commercial objective is to turn a technical requirement into usable, predictable capacity. IP quantity alone does not determine speed, completed tasks or acceptance by external systems. Those outcomes also depend on the destination, application and capacity of the contracted solution.

Operational needVLSM.io solution to assessPrimary criterion
Reach destinations that require IPv4IPv4 ProxyIPv4:Port ingress and IPv4 egress.
Use IPv6 for compatible destinationsIPv6 ProxyIPv4:Port ingress and IPv6 egress; validate every endpoint.
Retain or change the outbound originStatic, rotating or randomSelect for application sessions, intervals and requests.
Integrate address blocks into infrastructureIPv4 or IPv6 subnet rentalSize the prefix, delivery method, capacity and operations.
FROM CONNECTION TO RESULTTurn requirements into an observable operation.
Plan
Connect
Monitor
Destination planning, connectivity configuration, and result validation are complementary steps. The animation represents this workflow.
08

Turn scarcity into a capacity plan

Start with an inventory of destinations, supported families, applications, concurrent connections and fixed-origin requirements. Then separate workflows that must remain on IPv4 from those ready for IPv6. Compare solutions using the cost of the complete operation, including configuration, bandwidth, monitoring, support and continuity during failures. Avoid treating a generic per-IP price as the only measure of value. For automation, an interrupted session or a missed deadline may matter more than the difference between two addressing offers.

Bring that assessment to VLSM.io to size IPv4 Proxy, IPv6 Proxy or subnets for your project. Validate a representative workload before scaling, and track completed tasks, latency, failures and egress behavior. IPv4 exhaustion is a real technical constraint, but planning makes that dependency manageable: preserve the compatibility required today, use resources efficiently and introduce IPv6 where it already supports the operation. Combining technical understanding with suitable connectivity provides a foundation for consistent growth.

09

Institutional references

Reference documents on the topics covered in this article.