[qi:051] The time has finally come for the world to migrate to IPv6 from IPv4 -– or at least that was the message delivered by a collection of networking experts at the RIPE 55 conference late last month in Amsterdam. Out of this conference came a hilarious and very geeky song about how this change to IPv6 will more than likely cause operations and routing issues for network operators throughout the Internet. Regardless, it appears that the addressing scheme for every device on the Internet may finally be set for a transition — and the networking issues that may ensue could be far-reaching.
Taking a step back, it is absolutely clear that IPv6 offers significant networking advances, such as the ability to provide more addresses for devices on the Internet (3.4×1038 addresses total as compared to IPv4’s 4.2 billion address), an easier way for devices to autoconfigure their own addresses, a built-in mechanism for multicast and data security using IPsec. All of the IPv6 features promise to make the Internet scale better, support new services and have tighter security.
And while that is a good thing for the Internet, what is not so good is the probable pain of transition. After 20 years of building, running and fixing issues on networks running IPv4, moving onto IPv6 for a network operator is like ending of a relationship – painful but inevitable, and with the promise of meeting someone even better right around the corner.
As a simple example, according to a RIPE 55 presentation on the global state of IPv6, the number of prefixes (networks or portions of networks) running the new protocol is close to 1,000. Compare that to the number of global IPv4 prefixes — upwards of 200,000 and counting (and to the networking experts out there, I am aware that the addressing allocations and mechanisms in IPv6 allow for greater aggregation, thus reducing the overall number of global prefixes). Even with the number of prefixes on IPv6 an order of magnitude less on IPv4, there are routing issues that network operators will need to deal with on a daily basis. Those routing problems lead to packets that travel in loops, disappear into routing black holes and are hard for experts to diagnose because of the lack of operations tools and experience. And that is without the widespread use of IPv6 multihoming, the ability for an organization to use two different network operators for connectivity and not commit to a monogamous relationship. Like dating two different people, splitting your packets between two network operators can cause problems as well.
Even vendors that have had IPv6 support in their products for years, such as Juniper (JNPR), still see significant issues on their firewalls and need to resolve them. In fairness, all networking vendors have bugs in nearly every networking protocol — including IPv4 — but the lack of operational experience by the network operators makes debugging these issues harder and more time-consuming. It is clear to everyone that IPv6 is still in its infancy when it comes to global scale and operations. That translates into slower web sites, more downtime and fewer Google (GOOG) ads delivered.
So what is a typical organization to do? If you’ve been focusing on your web 2.0 application and ignoring the network as technology that just works, the time has come to learn about networking and IPv6. If you’re going to have a significant web presence in the next few years, you will want to use more than one network operator for Internet connectivity and that will undoubtedly result in new operational issues as IPv6 networks interconnect, the prefix counts increase and more multihoming gets put to use.
There is a new girl in town, her name is IPv6 and she’s clearly better than your old flame. Yet, like all relationships, there will be some bumps along the way and it will take some time for you to get used to each other. I suspect that we’ll be hearing a few new songs from the networking geeks in the near future, perhaps this time with a tune that you can dance to.