44/8

Tue Jul 23 14:47:34 UTC 2019

>I can guarantee you that Akamai is very much run by beancounters in addition to engineers. I have first hand experience with that.
>
>I can also assure you that it’s quite unlikely that any of Comcast, Netflix, Facebook, Google, AT&T, T-Mobile, or Verizon just to name a few of the biggest are managed without >due consideration of input from the bean counters. (I’d bet at each of those companies, the day that engineer beats beancounter in a disagreement is rare, indeed).
>
>Each and every one of those large companies has deployed IPv6. Some to a greater extent than others. Facebook and T-Mo stand out as the prime examples, having gone all->IPv6 in as much of their network as practicable today.
>
>The problem with the approach you are taking to IPv6 cost-benefit analysis is that your claim of no ROI doesn’t actually hold true.
>
>The cost savings from a full-on deployment of IPv6 and moving to IPv4 as a service at the edge can be significant. They are hard to capture without very good cost accounting >and the problem really tends to be that engineers are lousy cost-accountants and good cost accountants have a hard time understanding what IPv6 brings to the table.
>
>It’s also true that some fraction (though now diminishing) of the ROI from a v6 deployment cannot be realized until some other parties also deploy IPv6, but there’s good news >on that front, too… More and more of those parties are realizing the need to deploy IPv6.
>
>Owen

The common denominator for all of the companies listed is the size of their deployment.  The carriers needed to handle very large scale mobile networks that they could not possibly get a large enough allocation for.  The alternative CGN gear would have doubtless been extremely expensive as well.  They also have the engineering and financial horsepower to hold their suppliers to the fire to make all of the devices together well with v6.    Another advantage they have is that the lifespan of a mobile device and it's infrastructure is pretty short so they are not dealing with a lot of legacy devices.   It helped a lot that the v4 allocations were drying up at around the same time that the mobile networks were in full upgrade mode deploying 4G and LTE.

Facebook, Google, and Netflix deployed v6 mostly because there were so many mobile devices using it that it could not be ignored.  These are all market/financial forces at work, not some pioneering engineering drive.  In the corporate world we have to provide a reason to spend money so unless there is a business drive to deploy v6, it's not going to happen.  That pressure is mounting but not at a breaking point yet.  The two main pressures would be the cost of expanding into more v4 space and what your customers want.  The move to cloud based services means that the corporate demand for v4 space is actually decreasing since they are not hosting as many Internet facing applications as they once were.  They don't need to move to v6 since their cloud providers are doing it for them.  There is also a move in a lot of corporate networks to get away from dedicated circuits and use the Internet as transport.  As this happens, the corporate network does not even care that the service provider is using v6 transport to carry a tunnel.  V4 vs v6 becomes a non-issue over time and the pressure to change everything over to v6 goes away since the v4 space is not growing.

Steven Naslund
Chicago IL