Sunday update --- we're still in the small upward trend noted a fortnight ago, but it may be ending:
It's a bit early to say that the current rise in numbers is over, as today's 2,032 isn't far down from the recent maximum of 2,099 seen on 2018-09-12. Nevertheless, to the eye it looks like the end of a small peak. If that was the customer trial then it wasn't very extensive.
It also leaves us with a bit of a dilemma when trying to interpret the IPv6 activity of the last year. Customer trials are generally much more extensive than internal ones, yet the 2K peak of the past month's hypothetical "customer trial" is tiny compared to the three 8K-10K "internal trial" peaks observed over the last year. It doesn't look right at all.
I do see two possible scenarios that fit the observed IPv6 counts, but they're both a bit worrying. The common part of these scenarios is this: a year ago, VM started deploying DS-Lite internally in a Big Bang plan that started with adding IPv6 to all of its employee desktops and internal routing, since DS-Lite runs over internal IPv6 infrastructure. If so, then the first two peaks at 2017-09-28 and 2017-12-14 represent two phases of internal testing, and the slow rise to a third peak at 2018-03-29 represents gradual rollout of IPv6 routers all across the country. That internal test was terminated in the first fortnight of April 2018 for reasons unknown, but the obvious candidate for an explanation is that DS-Lite was discovered to be unfit for Virgin's internal use at this early stage, though still "adequate for customers" --- hence the current field trials now continuing with DS-Lite.
The trouble with that scenario is that DS-Lite needs internal IPv6 infrastructure over which to run, including IPv6 on employee desktops for support, yet we're not seeing APNIC counts up to 10K now like we did before --- the current numbers seem far too small. The above scenario does sort of work if the current (alleged) "customer trial" were of Dual Stack instead of DS-Lite (based on the premise that employees rejected DS-Lite for their internal use), since Dual Stack doesn't require extensive deployment of IPv6 across Virgin infrastructure, only in field equipment. The problem with that premise is that it's not what people suggest is happening.
An alternative scenario is that perhaps DS-Lite is being deployed to customers for trialing without any internal IPv6 deployment within Virgin Media at all because internal rollout is being left for later, after most of the Internet is IPv6. That could explain the numbers as well, since high-volume IPv6 trunks are separate pipes and the corporate admin network doesn't have to use them. However, this would seem to imply that they're not capitalizing on the main benefit of DS-Lite, namely moving to IPv6-only internally. It fits the APNIC data, but it wouldn't be saying a lot for VM's belief in IPv6 if they're not eating their own dogfood at this stage.
Morgaine.
