Transition strategy: dual stack, IPv6-only and IPv4-as-a-service
Compare the main transition strategies against cost, complexity and residual IPv4 dependency. Use current deployment-status data (RFC 9386) and the pros/cons of each technology (RFC 9313). Pick a strategy for a given organisation and segment.
Take the quiz · pass mark 70% · 9 questions
Explanation
There is no flag day. Every real network runs IPv6 alongside, or on top of, an IPv4 world that will not disappear for years. "Transition" is therefore about choosing, per segment, how IPv6 and IPv4 coexist until IPv4 can be retired. RFC 6540 (BCP 177) sets the baseline: all IP-capable nodes must support IPv6, so new equipment and software should never be IPv4-only.
The three strategy families
Dual stack. Every device runs IPv4 and IPv6 in parallel. It is the simplest to reason about and the most compatible, but it is the most expensive to operate: you maintain two of everything — two addressing plans, two routing tables, two firewall policies, two monitoring views — and you still need enough IPv4. It does not solve IPv4 scarcity; it postpones it.
IPv6-only with IPv4-as-a-service (IPv4aaS). The access network and hosts run IPv6 only, and IPv4 reachability is provided as a service at the edge: NAT64/DNS64 (L6.04) with 464XLAT (L6.05) for clients, or a carrier mechanism such as DS-Lite, MAP-E or MAP-T (L6.06). This removes IPv4 from the LAN, shrinks the operational surface, and is where large mobile and access networks have gone. The cost is translation infrastructure and a few IPv4-literal edge cases.
IPv6-mostly. A pragmatic middle ground on shared LANs: capable hosts move to IPv6-only (signalled by DHCPv4 option 108, L6.05) while legacy hosts keep getting IPv4 on the same segment. The network drains toward IPv6-only without a hard cutover.
RFC 9313 lays out the pros and cons of these IPv4aaS technologies side by side; RFC 9386 tracks real deployment status and shows the trend toward IPv6-only access with IPv4aaS.
Choosing
Score each segment on four axes:
- IPv4 dependency: how many IPv4-only services, appliances or literals must keep working?
- Scale: how many hosts, and is IPv4 address pressure real here?
- Control of the endpoints: can you set host policy (managed fleet) or not (BYOD, servers you do not own)?
- Operational cost: can the team run two stacks, or is one stack plus translation cheaper?
Rough guidance: data centres and server farms are often dual stack (you control them and want direct IPv4 and IPv6); large client access networks trend to IPv6-only + 464XLAT; mixed office LANs suit IPv6-mostly; greenfield should start IPv6-first and add IPv4aaS only where needed.
ictNetworks scenario
ictNetworks runs its data centre dual stack, its corporate Wi-Fi as IPv6-mostly (option 108 for managed laptops, IPv4 still offered for legacy gear), and its mobile/broadband access as IPv6-only with NAT64/DNS64 and 464XLAT. Its rule, from RFC 6540, is that nothing new is bought or shipped IPv4-only, so the dual-stack and translated segments keep shrinking toward IPv6-only over time.
Drawing diagram…
Worked examples
Example 1: The cost of dual stack
A site has an IPv4 firewall policy of 85 rules. Running the same site dual stack means writing and maintaining the equivalent IPv6 rules too, so the perimeter policy grows to 170 rules — plus two monitoring and logging views. This doubling of operational surface is the core argument for moving endpoints to a single stack where you can.
Example 2: Why IPv6-only scales
A single /64 access segment provides 2^64 = 18446744073709551616 addresses, so address pressure on the LAN disappears entirely. The only scarce resource left is public IPv4 at the translator, which is shared among many subscribers (L6.06).
Example 3: Sizing the plan
An organisation receives a /32 allocation. It can carve 2^(48-32) = 65536 site /48s from it, far more than it needs, so the addressing plan is never the constraint — the strategy choice is about IPv4 coexistence, not IPv6 space.
Lab
Goal: Score the reference topology's segments and recommend a strategy for each.
You need: the course reference topology (data centre, corporate LAN, guest Wi-Fi, mobile/broadband access), and the four scoring axes above.
- For each segment, record: number and type of IPv4-only dependencies, scale, endpoint control, and current operational cost.
- Score each segment and map it to dual stack, IPv6-only + IPv4aaS, or IPv6-mostly, citing RFC 9313's pros/cons for the IPv4aaS choice.
- Note, per segment, what must change to retire IPv4 (which appliances, literals or services block it), referencing RFC 6540 for procurement.
- Use RFC 9386 deployment data to justify the direction of travel.
- Produce a one-page per-segment recommendation and a sequencing plan (what moves first).
What to submit: the scoring table, the per-segment recommendation with justifications, and the sequencing plan.
Summary
- There is no flag day; choose a coexistence strategy per segment.
- The three families are dual stack, IPv6-only + IPv4aaS, and IPv6-mostly (RFC 9313).
- Dual stack is simplest but doubles operational cost and does not solve IPv4 scarcity.
- Score segments on IPv4 dependency, scale, endpoint control and cost; drain toward IPv6-only.
- RFC 6540 requires IPv6 support in all new nodes; RFC 9386 tracks the deployment trend.
- Next lesson (L6.02): tunnelling IPv6 over IPv4.
Quiz imported · 9 questions · pass mark 70%. Lab questions are not auto-marked.