I have spent a lot of time recently researching and working on large scale IPv6 deployments for ISP’s (Internet Service Providers). The business case here has been clear to me for quite some time; IPv4 addresses have all but run out and the scenarios for continuing to provide internet access (some flavour of address sharing) all cost money and potentially damage user experience. What has not been so clear to me however is what the business drivers for IPv6 deployment for content providers are.
Some of the Internet’s largest properties have been dual-stacked for some time: Facebook, Google and Netflix are all prime examples. But what made them deploy IPv6? From the perspective of a content provider, lack of IPv4 addresses is not such a pressing problem; after all, it takes far fewer IP addresses to host content at a large scale than it does to consume it. I’d always assumed it was a mix of Internet altruism and “best practice” rather than being driven by any hard need. That was until I started really thinking about what the Internet will look like in the coming years.
If ISP’s wish to continue to provide internet access to their customers they either need to come up with (usually, buy) more IPv4 addresses or start using some kind of transition technology. The latter is expensive and complicated while the former is both expensive and finite. It’s a fair conclusion that ultimately, service providers that are still growing, will need to deploy IPv4/v6 transition mechanisms to continue to provide access to the IPv4 Internet. Regardless of what service providers do with IPv6, this transition will be vital as customers still wish to access content that is only available via IPv4.
As there will not be enough IPv4 addresses to issue one per customer, some form of address sharing will be necessary. These technologies are typically grouped under CGN (Carrier Grade Network Address Translation) and require specialist hardware. This hardware does not come for free and it will be a cost service providers will have to bare. Exactly when depends on how fast they are growing and how many IPv4 addresses they have but eventually, most will have to do it. This technology adds more complexity and cost to a service provider and puts an extra obstacle between the customer and the content they are trying to access. For IPv4 only content providers, CGN poses a problem. This extra obstacle is outside of their control and sits directly between their customer and the content they are hosting.
Although CGN has been the norm in mobile networks for years, it is (currently) less common in fixed line ISP’s. The volume of traffic per customer is typically much higher in fixed line operators than in mobile and performance expectations are often different. People may be accepting of some buffering on their mobile phones however this becomes less acceptable when streaming high definition movies to afamily TV. Exactly how ISP’s implement and capacity plan CGN is yet to be seen but the fact remains – it will be another item that content providers will have to rely on service providers to implement well. Even putting aside potential problems with performance and reliability, address sharing has other undesirable side-effects related to security and analytics. If multiple customers appear to be connecting from the same IPv4 address this makes analytics harder and may trigger security systems on certain websites such as banking or gambling.
Of course, if the content provider is already dual-stacked, this doesn’t matter at all, the resource can be accessed natively via IPv6 without any concern for what the ISP is doing with CGN. And that, ultimately, is what I believe is the strongest driver for a content provider to offer their resources via IPv6 – User Experience. Offering a native IPv6 service removes obstacles between content and customers leaving fewer things to go wrong and a better overall experience for the consumer of that content.
We’re not yet at the stage where IPv4 address sharing among customers is common but it is not far away. The time for providers to ensure their content is dual-stacked is now. Otherwise they risk a degraded experience for their customers as IPv4 address exhaustion really starts to bite.
The views expressed in this article are entirely that of the author. They do not necessarily represent the views of the author’s employer(s) and/or their affiliates.