IPv6 support on Virgin media

[dgcarter]

Does anyone know whether (and if so when) Virgin plan to implement IPv6 on its network?

[jamesmacwhite]

From doing a bit of research this on myself. I believe the performance bottleneck may be due to the Intel Puma 6 SoC in the CPE. I believe the Hitron router uses this as well as the Super Hub 3 and the performance bottleneck is actually the modem. Which could explain a lot of things, given the Puma 6 SoC is infamous for issues with just about anything to do with latency and performance and not necessarily the network itself.

I've had a couple of VM residential customers who have a Hub4 confirm that they had better performance with 6in4 with a Hub4, so it might be the issue here. The Hub4 still uses Arris and the Puma chipset, but its the 7 series, so they seemed to have fixed or ironed out some of the bugs.

Further evidence from another ISP in Czech Republic shows very similar CPE hardware having the exact same problems here with 6in4, so it is now looking like CPE problem.

I don't know if there are plans to have a business version of the Hub4, whatever it might be.

[Anonymous]

Good luck and thanks! Having an account manager does have its advantages at times.

I did my bit back in the day (3-4 years ago) along with the IT manager at my work at the time. We had a substantial VM Business connection through the business park's fiber network. They were really not interested but we kept plugging away and the issue was escalated. Sadly we both moved on in our careers before anything happened...

[Anonymous]

---

@jamesmacwhite wrote:

From doing a bit of research this on myself. I believe the performance bottleneck may be due to the Intel Puma 6 SoC in the CPE.

---

Don't believe that is the case. I had experience of the issue with an SH1 and my current SH3 (both in modem mode). These devices have quite different architectures and the SH1 didn't used to have problems and (AFAIK) the firmware version didn't change. It looks to be something deeper in the VM network doing the throttling/de-prioritisation.

[jamesmacwhite]

It is possible it is deeper than that, but if there is throttling, Virgin Media refuse to acknowledge it and have continually denied it for years. So taking that at face value, going on the basis they aren't in fact doing this, that makes me look at other potential elements, but of course it is only guess work.

The cases of other reports of the same issues with 6in4 all seem to point back to the Intel Puma 6 SoC. Equally, from a few customers that tested it, the change of modem to Hub4 seemed to improve things, so while it might not be the only problem, why the performance difference? If if was the network you'd expect that to be constant. The main changes with the Hub4 would be the Intel Puma 7 SoC and DOCSIS 3.1, although I don't see how DOCSIS is really related to this specific issue.

Perhaps there is a CPU bound requirement for 6in4. The SH1 and SH2 were Netgear if I recall, then the SH3 and newer Hub4 are both Arris Intel Puma SoC. Typically ISP provided kit is under powered or "just enough" to the basic job as per a contract, so there may be merit to it, but who knows.

[ChrisJenkins]

It seems unlikely that the SH3/Hitron hardware/firmware is an issue in my case because:

1.   In modem mode (which is how I run) the SH3 (or in my case the Hitron) should just be dealing with the minimum needed to allow packets to flow. It certainly should not be doing any protocol related processing and so all IPv4 traffic should be treated the same by the SH3/Hitron.

2.   If I create a VPN connection to my VPN provider (over VM IPv4 of course) and then create a 6in4 tunnel over that then I can get much, much higher IPv6 speeds (over 100 Mbit/s, despite the significant overhead of the OpenVPN VPN). In this case the VM network only 'sees' a VPN connection and does not 'see' the 6in4 traffic.

This second point seems quite definitive to me. I have tried with several different VPN endpoints in the UK and elsewhere and also several different tunnel servers and all show this far higher throughput when the VM network cannot see that the traffic is protocol 41.

[jamesmacwhite]

I do agree with you. I have also tested a IPv6 enabled VPN (Mullvad) and got much better speeds. If the 6in4 tunnel terminates on a VM connection the speed is horrible if it is encapsulated over UDP i.e. Wireguard or OpenVPN then the performance is much better, so yes you'd think it is VM throttling, but I then don't understand why they'd deny it all this time or why customers with the Hub4 saw much better performance, without doing anything other than swapping the CPE with different hardware specification, doesn't make sense and contradicts both theories.

Even their traffic management page says nothing, of course they define that so they can say whatever they like and even perhaps hide behind the "other traffic measures", which could be anything. Maybe a formal complaint to the higher authority or an FOI request on specific treatment on 6in4 traffic in their network is one way to confirm it.

It makes me think there are multiple parts to this and it's not just straight throttling of a protocol, you'd like to think if it was it could have been fixed maybe several years ago. This problem isn't exactly secret and is fairly well known across different communities, even if it is niche.

Given you're a VM Business customer you'll at least have some form of SLA response, so maybe you'll have more luck. Best of luck!

[matthewsteeples]

---

@ChrisJenkins wrote:

This second point seems quite definitive to me...

---

That would also be explained perfectly by the hub theory though, because all the hub would see is the VPN connection and it wouldn't be able to tell that protocol 41 is in use or need to do any processing to cater for it

[ChrisJenkins]

That would be true if the hub is in router mode, but in modem mode it does not 'look' at the packets nor do any processing (if it does then that is a serious bug).

Do we have any links to the evidence that much faster speeds are obtained using an SH4? Do we know for sure if the SH4 was in router mode or modem mode for those cases? I'd like to review exactly what has been claimed.

[jamesmacwhite]

In modem mode the CPU will still be present though. You remove most of the routing functionality of course so easing the load, but the CPU is still required for the WAN side. It is why the latency issues with the SH3 in modem mode are still present, if it is a hardware issue from the CPE and can't be fixed with software or firmware. Modem mode would help, but it can't fix the underlying hardware level issue.

From the few people I talked to outside of this community forum who have a Hub4 and known what 6in4 is, they were using modem mode. The only way to configure a 6in4 tunnel in router mode is then having to NAT a LAN client for protocol 41 and deploy the tunnel on that client, so no one will really be doing that I don't think.

Given the Hub4 is newer and the not many people will be doing 6in4 I admit, I cannot verify the claim myself. I'm not currently in a DOCSIS 3.1 area, so I doubt I'd be able to convince Virgin Media to give me a Hub4, I'd even consider paying for it, if it was an option just to see for myself. I guess the alternative is throw out a general "All Hub4 users using a 6in4 tunnel" type of message to see what the story is. Of course, a few people isn't a relative sample.

The latest known info I'm aware of on the Hub4 is, Virgin Media are offering it outside of the Gig1 service, but only to selected areas, likely where DOCSIS 3.1 is available, so they can utilise the extra bond channels to reduce network load, or use it as a "WiFi upgrade" for the home. Eventually it will become widely available, but I don't think it will happen just yet.

[ChrisJenkins]

James, this doesn't really make sense... Why would the hub, in modem mode, have to spend much more CPU processing protocol 41 packets than other packets (given that it should simply be 'passing through' all IPv4 packets in the same way)? If there is a general issue with the CPU power of the hub (in modem mode) then it would be impacting all traffic not just protocol 41 traffic. I pay a great deal of attention to the performance of my Internet connection (throughout and latency). I have monitoring in place that measures the performance every 15 minutes and I have data from this going back for over 3 years (SH2, then SH3 now Hitron). Other than occasional periods when there has been an actual service issue (i.e. VM network problem) I have not observed any major issues with latency or throughput (apart from the dreadful IPv6 throughput of course).

Certainly the hub processing power might be an issue in router mode *if* the hub was for some reason doing a lot of extra processing for packets carrying protocol 41 (it should not be of course but who knows what the firmware does) but again in modem mode the doesn't make sense. We're not talking about a small impact here; we are talking about the difference between 500 Mbit/s and 20 Mbit/s. I cannot conceive that this is due to a lack of CPU power in the hub when the hub should not even be involved (other than simply passing the packets through uninspected).

If we can see some real evidence that SH4 make a big difference (in mode mode) then I'll happily review my opinion but at the moment it seems like conjecture to me.