IPv6 support on Virgin media
Does anyone know whether (and if so when) Virgin plan to implement IPv6 on its network?
Forum Discussion
I've always wondered how the numbers on that page are worked out. In terms of total number of customers BT is the biggest ISP, followed by Sky, then Virgin is a distant third due to the limited availability of cable services. But the "users" on that website show Virgin has the most users, followed by Sky then BT.
The question of what the APNIC IPv6 counts actually represent is asked quite regularly, so I'll just paste what I wrote a year or so ago in a discussion at the UK IPv6 Council group, which is hosted at LinkedIn. BT had just reached its first 1 million daily IPv6 "counts" as seen by APNIC and was congratulated for it, which led to the question being raised:
There's no need to guess what the numbers mean, as APNIC's excellent George Michaelson regularly gives talks about the measurements they make. For example, "Weighting the World one Click at a Time" was given at the Cisco-Ecole Polytechnique symposium not long ago.
As scientists and research engineers are well aware, sometimes you can't measure the thing that you want directly, so you use a "proxy" from which to estimate it indirectly. (Not the same thing as "proxy" in networking, but nevertheless related.) George gave the rather nice example of Cavendish trying to determine the mass of the Earth, by way of introduction.
In our case, the numbers we want cannot be measured directly either, and ISPs won't generally reveal anything under the excuse of commercial sensitivity, particularly if we want updates every single day. And so we use proxies from which to derive estimated figures because we cannot obtain official ones.
APNIC's measurements are proxies.
Although they do not reflect anything as precise as the number of actual IPv6 users per ISP, they are nevertheless much better than nothing at all, and indeed highly useful both as a proxy and in their own right to show trends and changes in rate of growth. Even the ISPs themselves cite APNIC figures, incongruously explaining that they are prevented from releasing official ones, as we have seen ourselves at UK IPv6 Council meetings. :-)
APNIC's measurements also have another good property, in that they are gathered automatically and hence any artifacts or skews apply equally to everyone. While the numbers are somewhat artificial in absolute terms, this is public information and is well known. The "1 million" marker was of course of no special significance, but it's as good as any other, and it did at least give us the opportunity to pat a provider on the back for making good progress in IPv6 deployment, just like awarding plaques does. :-)
Numbers are always welcome, even proxies.
Exactly the same applies to Virgin Media as applied back then in respect of BT. Virgin management doesn't release publicly the data that we need to quantify its degree of IPv6 deployment, so we use proxies instead. APNIC's stats are one of the most informative proxies, but there are many other sources available too such as Cisco's 6Lab and Google, and all of the other RiRs provide IPv6 stats as well, not just APNIC.
For hard details on APNIC's measurement methodology, I highly recommend searching for "George Michaelson" in all the usual places.
Morgaine.
Just a quick graphic update to confirm that Virgin IPv6 counts are still in their collapsed state, post 2018-04-10:
Hopefully they're now sorting out the final details prior to launch of IPv6, after what appeared to be two significant IPv6 trial periods and then 2 months of infrastructure rollout. Yes yes, I know, too optimistic by far. :P
Morgaine.
Thanks for the update, I've been refreshing this page regularly to keep up to date
You can “subscribe” to messages
Sunday update. :P
In short, we're still in the post-2018-04-10 collapsed state of Virgin's public IPv6 activity, so my tentative interpretation since that time continues to seem possible.
Morgaine.
I really hope you're right, but based on recent experience with other trials, I would imagine that IPv6 rollout is going to start on an opt-in basis with people registered for trials, and then ramp up from there (as they did with the Intel bug fix). I don't think we're going to get a "big bang" approach.
@matthewsteeples: Although it's only a guess, if my interpretation of the numbers is correct then the rollout on an opt-in basis for trialists has already been carried out.
Of the 5 vertical red bars on my graph, bar #1 marks the peak of the initial convening and acceptance of IPv6 trialists under NDA, with a fast but rather erratic pattern of growth because recruitment spreads by word of mouth. Bar #2 marks the peak of the 2nd trial period, but this one has a very different curve to the first because the trial group already exists at the moment when Virgin announces that new testing is required and makes a request for feedback. The reaction is nearly instantaneous by those trialists who are available and checking their notifications.
Between bars #3 and #4 lies what seems to be a 2-month period of IPv6 infrastructure rollout, a very steady rate of increase which is a natural outcome of hardware deployment performed by a fixed size team. The opt-in happened earlier when trialists self-selected, probably picking up a "live rollout test" session soon after the new equipment was deployed in their area, merely by rebooting their CPEs. There is some indication that the rollout actually started considerably earlier than at bar #3, because the drop-off in APNIC counts from the peak at #2 does not have the expected reverse exponential curve. This suggests that we are seeing a superposition of two components instead, one the expected drop-off from the original centralized test facility, and two, a new contribution arising from CPEs using the new IPv6 infrastructure as it is rolled out slowly across the UK, if their CPEs are enabled for it.
The peak in IPv6 counts by the end of rollout at bar #4 lies within the range of the two trial peaks, which is additional supporting evidence that the same set of people are involved, the trialists. So, I think we're there already. :)
The icing on the information cake comes from the extremely abrupt drop in counts between bars #4 and #5. By turning off trialists' IPv6 access, Virgin is treading dangerous ground, because this will be annoying them immensely, especially the gamers. This is why I think Virgin are intending to launch an IPv6 service imminently, to avoid major discontent among enthusiasts who won't put up with it for long.
Of course it's all conjecture, but it does seem to be consistent with the APNIC data which is statistically very robust. In any case, it's all we have, as Virgin considers IPv6 deployment a secret. I just can't get my head around the incomprehensible pointlessness of that. It has lost them customers.
Anyway, I think we'll see soon enough. :)
Morgaine.
Some general UK IPv6 news for today, 2nd May 2018:
BT has just reached another IPv6 milestone --- 4 million daily IPv6 counts as measured by APNIC. Congratulations, BT !!! :)
Here are the last few daily samples taken from APNIC's public stats:
DATE AS Users IPv6-Users %UKv6
========== == ========== ========== =====
2018_04_29: BT 11,411,215 3,978,234 23.32
2018_04_30: BT 11,422,473 3,996,894 23.32
2018_05_01: BT 11,401,525 3,999,437 23.23
2018_05_02: BT 11,407,197 4,006,326 23.25 <-- Ding!The graph below shows BT's IPv6 growth (light blue curve) in the context of the other "Big Three" UK ISPs, with Sky far in the lead:
4 million is 34.7% of APNIC counts measured over BT's AS, so achieving it deserves a big pat on the back. When one third of your traffic sessions are carried over IPv6, it's clear that IPv6 has become a solid pillar of the current Internet, not just the Internet of the future.
The graph also highlights a less happy observation, namely that at BT's current rate of IPv6 growth, it will be many years before all BT Broadband customers have an IPv6-enabled CPE --- that's a major problem which needs to be addressed by them. Hopefully the company will roll out their HH6 CPE quickly to all customers in response to their HH5 upgrade issues, or find some other way of overcoming the problem.
If I have interpreted the APNIC data for Virgin Media correctly, BT are about to find themselves in third place behind both Sky and Virgin, and a distant third at that, because Virgin implemented a mandatory CPE upgrade plan recently so it's likely that most remaining legacy kit can support whatever is coming. If VM manages to provide a large majority of its subscribers with IPv6, BT is going to have to engage an emergency action plan or suffer years of embarrassment and a costly loss of customers.
Morgaine.
Wonder why the BT numbers started ticking up recently having sort of stagnated for a while. I thought it might be because they started a CPE upgrade program of some nature.
The other interesting development in non-VM IPv6 is that EE seems to be growing too. Now up to 8.87% of their users.
I'm not sure apocalyptic projections on BT's in ability to deliver IPv6 to their customers will amount to much! If in the unlikely event they start to lose customers money will no doubt be found for a CPE upgrade for HH5 users. With a general move to FTTP and G.Fast happening there will be a significant natural churn in their CPE base over the coming years anyway.
I do wish Virgin would break cover and be a bit more open about their deployment strategy and timeline. Seems that if they are close to pushing the go button that there's no earthly need for secrecy.
Virgin operate a policy where they say as little as possible, erring on the side of caution.
@davefiddes wrote:
Wonder why the BT numbers started ticking up recently having sort of stagnated for a while. I thought it might be because they started a CPE upgrade program of some nature.
No, nothing like that. The flat part of BT's growth curve arose when their network developed a fault under which IPv6 provisioning would disappear from individual lines and never come back, ever. Unfortunately this engineering issue was turned expertly by BT into farcical comedy because IPv6 had been rolled out to their network without ever telling anyone in support, business management, customer information systems / website developers, and nor was IPv6 added to BT Fault Reporting. As a result, nobody in BT could be approached for a remedy because the response was always "What is IPv6?", or alternatively, "BT doesn't provide residential IPv6". Totally surreal. (But thank goodness for the hard data from APNIC.)
Eventually the BT engineering division was informed by BT Community Forum staff that many customers were reporting this IPv6 vanishing act. I had been reporting falling APNIC numbers to them for a while so that was probably also supplied as evidence. After 3 months or so a fix was rumoured but it wasn't until a good 6 months elapsed overall before BT IPv6 finally came out of its plateau and returned to growth. No official explanation of the problem was ever provided, because after all, you can't explain officially the non-working of a service that officially isn't being provided. :P
They have some really severe problems in BT, and it's not technology that is letting them down. (After a very curious exchange in the UK IPv6 Council group recently, I now have some hope that BT engineers have awakened to the problem they created and are now talking to other BT divisions about remedies.)
Morgaine.
- which is the way it should be for any company. never commit.
That mention of turning off the IPv6 of gamer trailists brought to my mind the other (probably 2 year old) Elephant in the room being the nasty Intel Puma6 issue. Basically, the core of the SubparHub3 has a flaw which means that under even moderate load it is not capable of responding to aspects of the data passing through it fast enough and the packet latency soars into the 100s of milliseconds. This is bad news for real-time gaming as it means that timing-important game data gets held up enough to make some games either unplayable or to cripple the VM customer player relative to other players on ISPs who do not have the Intel device at the core of their CPE {See this topic - but it is now at 374 pages https://community.virginmedia.com/t5/Gaming-Support/Hub-3-Compal-CH7465-LG-TG2492LG-and-CGNV4-Latency-Cause/td-p/3271492/page/375 }. There had not been any indications that VM are in a position to update their way out of it (although a recent post looked a bit promising) - so perhaps they are holding out on switching on IPv6 because running a Dual-Stack on the Hub will make things even worse and they are going to hold out until Intel compensates them {here's a form that VM might want to consider filling out: https://www.classactionlawyers.com/puma6/ !} or they can get their hands on kit that can do the work as a SuperHub4...
I mean, take a look at my Max Latency (yellow graph) right now:
After all this secrecy and time in silence, the least that VM can do is make a small and very dramatic video of when they finally use one of these buttons to deploy IPv6 to production.
Off-topic Ah ha, I remember that Button - I got one for a Christmas present for my Silver Surfer mother. I see that an enterprising chap has produced updated software for (in the past Win7) and now Windows 10 - see: http://www.johnbruin.net/index.php/2017/07/22/newest-software-for-your-usb-panic-button-windows-10-compatible/ ...!
@SlySven: From your link:
"This alternative software for the USB Panic Button from Jam has stopped working in Windows 10 so it is finally time for a new version!
Big Red Panic Buttons need an Ethernet connection and IPv6 address! :P (This would also provide immunity from Microsoft's regular cycles of destruction.)
Just a Sunday update, no change since the 2018-04-10 collapse of Virgin's public IPv6 activity:
It's been over a month since that collapse, and while my interpretation still fits the data, something is beginning to feel wrong. Turning off the IPv6 access of trialists for this long would be quite a kick in the teeth after giving them a taste of IPv6 for over a year. It doesn't sounds right, even for Virgin.
Unfortunately, the only other explanation that comes to mind is a bleak one, that Virgin has called off IPv6 yet again. I hope not. That would be a disaster not only for the trialists but for the entire UK.
Morgaine.
The aim of trials is to try stuff out in the real world that you can't test internally or in a lab. Perhaps they found out that DS-Lite really does suck.:smileywink:
I think the DS-Lite policy was for all of the Liberty Global systems. Now that they are probably selling off all of the cable networks in Europe except for the UK and Eire perhaps the VM arm has had a rethink.
Hopefully they're rethinking as DS Lite sounds like the worst possible option.
https://support-en.upc.ch/app/answers/detail/a_id/1259/kw/IPv6
This is Liberty Global's Swiss service (not being sold) and according to the information on their website, you can't use "bridge mode" (Modem Mode) on IPv6. If that applies to VM I think a lot of people will be quite happy to stick with IPv4!
cje85 wrote:Hopefully they're rethinking as DS Lite sounds like the worst possible option.
https://support-en.upc.ch/app/answers/detail/a_id/1259/kw/IPv6
This is Liberty Global's Swiss service (not being sold) and according to the information on their website, you can't use "bridge mode" (Modem Mode) on IPv6. If that applies to VM I think a lot of people will be quite happy to stick with IPv4!
ooh, I'd really like to know the answer to the point about bridge mode. I've not used my virgin router as a router in about 7 years as I have a pfsense router behind it that I vastly prefer. I also want IPv6, but I was assuming that the router would also hand over all IPv6 info to my router to deal with too...
That document linked above (within Liberty Global's Swiss service) is expressed poorly and may lead to confusion, especially this part:
Due to the absence of the NAT server, the following functions and settings are no longer possible with IPv6:
Bridge mode
UPnP
Port forwarding / forwarding
Port triggering
DMZ hosting
With IPv4, the above-mentioned functions and settings can be configured directly on the modem.That's putting the blame on IPv6, when it has nothing at all to do with IPv6 but with IPv4's remote implementation in DS-Lite. Even worse, it seems to be suggesting that the "solution" is to get rid of IPv6 (although I may be misinterpreting that.) I'd rephrase the above more accurately like this:
Due to the absence of IPv4 NAT within the CPE under DS-Lite, the following IPv4 functions and settings in the CPE are no longer possible for IPv4 (although they still have equivalents in IPv6):
CPE LAN Bridging for local IPv4 devices (this has nothing to do with modem-only mode)
UPnP
IPv4 Port forwarding / forwarding
IPv4 Port triggering
IPv4 DMZ hosting
With IPv4 implemented locally on the CPE instead of over DS-Lite, the above-mentioned IPv4 functions and settings can be configured directly on the CPE. What's more, there is no need to disable native IPv6 when IPv4 reverts to local.
All of the above continues to be true when CPE functionality is split into modem-only operation plus a separate router.Unfortunately "the Internet" at large has confused bridge operation with modem-only operation way beyond repair, despite them having very different purposes and working entirely differently. A bridge merges two or more separate MAC address spaces into one so that you don't need to route between them at a layer above --- this "flattening" makes it easier for wifi access points or other devices to work with broadband. Modem-only operation of a CPE has the purpose of turning the CPE into nothing more than an interface on another router, typically a more featureful model, and is very different from running the CPE as a combined modem/router and then bridging its LAN to other local LANs.
Alas it's far too late to sweep back the tide of Internet misinformation, but at least we can be clear about the distinction here. More relevantly, there is no a-priori reason why modem-only operation cannot be used in a DS-Lite setting, although it does give the router more work to do since it needs to set up the IPv6 tunnel to the AFTR IPv4 gateway itself. It's nothing particularly onerous though.
Morgaine.
Sunday update, no significant change:
No news is bad news. My usual optimism is plummeting, this is not where I expected us to be this summer. :(
Morgaine.
I cleared out this graph repost, as @davefiddes explains below why it appeared broken initially.
Yep. It happens every week when you post. I think that images have to go through a moderator for a manual check. They always show up eventually.
Many thanks, @davefiddes, that explains it perfectly. (The image in my repost faired no better, so I cleared it out.)
I'm surprised that I hadn't noticed before. It seems that Lithium doesn't show posters the same thing that everyone else sees, at least for images until they get approved. I suppose it's a necessary safeguard on the Internet of today, but it adds to the support burden unfortunately.
