Igalia built a high-performance, open-source implementation of the border router component of the lightweight-4-over-6 IPv6 transition technology.

Lightweight 4-over-6: a compelling IPv4+IPv6 architecture

When you go to deploy a modern network, you’ll need to make both an IPv4 network and an IPv6 network, to give your customers access to existing IPv4 systems as well as to newer IPv6-only services.

But it’s a cost to have to build two networks. What if there were another way? With lightweight 4-over-6, there is.

Diagram of lw4o6 network, showing tunnels originating at CPE, traversing the operator's IPv6-only network, terminating at the lwAFTR

A network that deploys lightweight-4-over-6 is an IPv6-only network. The carrier’s internal network doesn’t have to route any IPv4 packets, saving costs and complication.

When a customer wants to access an IPv4 address, the customer-premise equipment (CPE), for example their home router, will tunnel that traffic in IPv6 and direct it to the “address family transition router” (AFTR). The AFTR, sometimes called the lightweight AFTR or lwAFTR, is on the other side of the carrier’s network and is connected to the IPv4 internet.

Compared to a technology like Dual-Stack Lite, Lightweight 4-over-6 has the advantage of being constructed in such a way not to require per-flow tracking on the AFTR side. The NAT functionality is pushed out to the CPE, not centralized into the AFTR. This allows seamless scaleout and failover via ECMP and other mechanisms; it doesn’t matter lwAFTR instance in a cluster processes a given packet in a flow.

Lightweight 4-over-6 has been a cornerstone of Deutsche Telekom’s next-generation Terastream deployments, allowing free failover and redundancy in the internal network.

IPv4 address sharing yields significant savings

But even if you haven’t already decided on an IPv6-only internal network, lightweight 4-over-6 has another compelling advantage: it allows end users to be mapped to slices of public IPv4 addresses. User A might get IPv4 address 1.2.3.4, but only ports 1024 to 2047; user B can be on that same address, but have ports 2048 to 3071; and so on. Given the cost of IPv4 addresses these days, economising on total IPv4 allocations is an attractive proposition.

For more information on Lightweight 4-over-6 in practice, see OTE engineer Kostas Zorbadelos’ presentation at RIPE76.

Snabb lwAFTR

The Snabb lwAFTR is an open-source implementation of the AFTR component of a lightweight 4-over-6 deployment. Developed entirely by Igalia, the Snabb lwAFTR is the most cost-effective fast lwAFTR solution on the market, even winning open calls-for-proposals involving established vendors.

Interesting aspects of the Snabb lwAFTR include:

  • Horizontal scalability by devoting multiple CPU cores to servicing traffic from a single Ethernet port.

  • Multiple-port horizontal scalability with multiple instances, all managed by a single management end-point.

  • CPU overhead in proportion to port count. A usual deployment will need two CPU cores per Ethernet port, plus one additional core for management tasks.

  • Support for binding tables of effectively unlimited size. The Snabb lwAFTR has been tested for binding tables containing up to 40 million entries (corresponding to 40 million distinct customer sites).

  • Easy management via configuration and state expressed in terms of the industry-standard ietf-softwire-br YANG model.

  • Low-overhead run-time binding table adjustments (thousands of binding table additions or removals per second without packet drops).

  • Open source, royalty-free licensing.

Igalia and lightweight 4-over-6

Igalia is the sole developer of the Snabb lwAFTR, with support from Deutsche Telekom. The Snabb lwAFTR has also been combined with Juniper’s vMX control and management plane, and is purchasable as a product directly from Juniper.

Igalia is happy to offer support and consultancy related to the Snabb lightweight 4-over-6 AFTR. Contact us today to be put in touch with one of our colleagues.