IPv6 Routing
The Next Frontier
by Bill Kine, Product Manager, Spirent Communications

IPv6 is here. The communications industry no longer refers to IPv6 in the future tense, discussing how wonderful it will be. IPv6 has arrived, and it is fulfilling all of its promises. IPv6 products, applications and solutions are available today. IPv6 networks are gradually increasing in size and popularity worldwide. Hosts, servers and routers are routinely using 128-bit addresses, neighbor discovery and auto-configuration. Therefore it is now time to advance the discussions of IPv6 to the next logical level - routing.

Routing is a fundamental part of the Internet Protocol, regardless of the version number. In fact, IP was built based upon the concepts of subnets and routing. The same level of forethought and planning has gone into IPv6. Subnets, aggregation and even multicast routing can be easily denoted by the inherent IPv6 address and/or prefix.

The move to IPv6 was designed to be an evolutionary process. Modern workstations, servers and routers should now be upgradeable to the next generation of the Internet Protocol. In fact, the whole upgrade procedure should be painless for the users. However, this migration will be a lengthy process and will require both versions to peacefully coexist for many years - perhaps an entire decade!

Routing protocols also need to adapt to the new version of IP. The four most common protocols: BGP, IS-IS, RIP and OSPF already have IPv6 extensions. In fact, all four of these protocols can support IPv4 and IPv6 concurrently, facilitating multiprotocol transitional networks. Most major router vendors have already implemented these extensions, known as BGP4+, IS-ISv6, RIPng and OSPFv3. These implementations build upon the vendors' solid IPv4 protocol foundations.

IPv6 Routing - The Brave New World

Although all of the major router and switch vendors have successfully implemented IPv6 routing protocols, actual real-world experience with these solutions is quite limited. IPv4 has had 20 years to shake-out (or work around) all of its bugs, issues and limitations. Over that time period, IPv4 and all of its ancillary protocols have been revised, optimized and enhanced many times. Furthermore, a great deal of collective and individual expertise has been accumulated regarding IPv4 networks.

The current suite of routing protocols was constructed specifically for IPv4 networks. Its associated options, methodologies and algorithms have been refined over many years to accommodate all of the nuances of IPv4. These protocols were all completed long before IPv6 was even conceived. In fact, although this may seem rather harsh, it is fair to say that IPv6 was an afterthought for these protocols.

There is a severe shortage of real-world experience with large IPv6 networks. The largest contemporary IPv6 network contains fewer than 1,000 routes - by contrast, the IPv4 Internet has 140,000 routes. Many of the fundamental principles of the IPv4 routing protocols remain unproven in the emerging IPv6 world. Maximum quantities of peer routers, route table sizes, convergence times and filter capacities all vary considerably as the users migrate to IPv6; in fact, many of these metrics are entirely unknown. Furthermore, the performance and scalability of routers and protocols in a mixed IPv4/IPv6 environment is even more unpredictable; and transitional networks are expected to exist for many years to come.

IPv6 Routing - The Next Phase of Testing

Over the past few years, users have successfully tested the basic performance and functionality of their workstations and applications over IPv6. These have typically been small-scale "proof of concept" tests in controlled laboratory environments. It is now time to expand these tests to large-scale Internet operations. Network architects should fully stress test their equipment and protocols in order to understand their performance and functional characteristics and limitations.

Testing will help determine how a modified IPv4 routing protocol operates in an IPv6 environment. Comprehensive functional testing will further validate that all of the IPv4 attributes have been adequately adapted for IPv6. However, additional routing protocol modifications may be necessary as new IPv6 features are developed.

Scalability tests are even more critical. Routing protocols have been designed to propagate 32-bit IPv4 addresses. Therefore their performance and scalability will be different for the new 128-bit addresses. Furthermore, IPv6 has been designed to overcome the address depletion problems currently associated with IPv4. And IPv6 fixes this issue in a big way! Routing tables could potentially contain millions of prefixes. (Or even billions or trillions or more!) This will directly impact the routing and protocol performance, and may even overload some systems. Emulating and testing the limits and the performance of the modified routing protocols such as OSPFv3 in huge IPv6 networks will help users design and optimize their IPv6 infrastructures.

Crystal Ball - What Will the Future Bring?

The current popular sets of IPv4 routing protocols have enjoyed many years of success. They have been modified to accommodate the new demands of IPv6, and they will probably successfully support this evolution of the Internet for several years. However, as the IPv6 Internet continues to expand and new features are developed, it is likely that the world will outgrow these legacy protocols. It is also likely that entirely new routing protocols will be developed specifically for IPv6. In the mean time, it is critical that equipment manufacturers, network architects and even end users fully recognize the limits associated with the current protocols and devices. In the past, users have tested the basic IPv6 functionality. Now is the time to take the next step and test the routing protocols of tomorrow's Internet.