Virgin Media Discussion Thread

Soldato
Joined
29 Dec 2002
Posts
7,176
We're now 1Gig enabled and was going to upgrade and take advantage of a Volt sim, until I checked and found GifGaff (who use the O2 network) are still cheaper (15Gb for £10 a month).

Just the upgrade for me, along with a wifi mesh refresh which will mean I can finally getting rid of the powerline kit that I've been using to make up for the wifi failings in the house. Asus and Netgear Orbi currently in the running....

You seem to have missed my earlier post, £8/m for 15GB and doubles to 30GB with £13 cachback using Quidco on MSM.
 
Associate
Joined
6 May 2011
Posts
1,380
Location
Inside the M25
That's a 12 month contract though, rather than month by month?

It's a sim for my son's phone - need to be able to cancel it or move tarriffs depending on whether he has done his chores and homework! (Yes, I am an evil parent)
 
Soldato
Joined
18 Aug 2007
Posts
9,688
Location
Liverpool
Great, another router distro to re-visit :D Logically I suppose a router specific distro should be optimised from he ground up with that role in mind vs a vanilla BSD install. Historically I tended to favour DDWRT/Tomato over OpenWRT, I notice OWRT has a VM image which could be useful for HA, though even with ESXi's direct pass through, I think i'd prefer bare metal and i'm likely thinking way too far ahead here. Also what happened to the i350 v3? I thought it was only me slumming it with fewer queues :D

Just noticed VM have started the BF sales push - if you can call it that. £62 for Gig1 as a standalone, i'm currently paying £59.99 for Gig1, 2 SIM's, basic TV and phone. It would be £55.99 if I dropped the extra SIM :rolleyes:

Yeah the image is tiny and very flexible. I tested it in VMWare before imaging it to bare metal to test. I didn't want to hose my OpenBSD router in case... well, in case... so it's currently the only thing installed and running on an Optiplex 7010 (Core i7 3770, 8GB RAM, 1TB Samsung 850 PRO). Not wasteful at all! :D

Now to decide whether to copy over the config to the original little box (two clicks in OpenWrt), or finally install something like Unraid or Proxmox onto the Dell (I've never played with them before despite the itch), and have OpenWrt running in one VM and use the other 0.9TB of the Pro SSD for more interesting things...

As for the NIC, I wanted the v3 but late night ordering and possibly some alcohol means I ended up with a T2-v2 as it transpires. I didn't even realise until I installed it yesterday haha! Eight queues per port will have to do for now. :p

For anyone else, I fixed the Unbound issue. Now I have dnsmsaq serving dhcp and local DNS resolution for hostnames, and Unbound doing DNS over TLS to my AdGuard Home adblocking server upstream:

50DwFIk.png

OpenWrt actually also has the AGH package available itself, but I already have the VPS running and I'd rather have DNS off prem in case of issues, as our mobile devices rely on it off-network too. It was fairly easy in the end.

I downloaded and left seeding an Ubuntu 21.04 torrent to test latency this morning, and my BQM didn't even flinch. My base and average latency is down compared to before, no yellow spikes at all, and my upload is responsive throughout. Cake really rocks. Just a tiny increase in average latency during the seeding, and back down again once I closed the torrent client (after ratio >2). I've cropped it because it's such a small increase during seeding you'd struggle to see it on a full graph. It's around 5am-7am (notice the extra latency in the blue band). Tiny! Before I'd have had wide blue and massive yellow spikes. I can't recommend cake and proper SQM highly enough - now I know why Dave gets so excited, it's transformative to the line, and no more hiccups in WiFi calling or whatever when someone is saturating the upload.

7rojqW4.png
 
Soldato
Joined
29 Dec 2002
Posts
7,176
That's a 12 month contract though, rather than month by month?

It's a sim for my son's phone - need to be able to cancel it or move tarriffs depending on whether he has done his chores and homework! (Yes, I am an evil parent)

Then you either need to pay the 30 day rolling contract rate (and accept that if/when you pull it, VM are likely do the same to your Gig1 upgrade) or find another candidate to use the SIM. What GiffGaff charge is irrelevant, they may use O2 as an MVNO, but they aren't O2 so you wouldn't qualify, neither will an O2 PAYG SIM.

To get your Volt benefits, you need to be a customer of both O2 mobile and Virgin Media broadband. Your O2 Pay Monthly mobile plan must be registered in the same name and address as your Virgin Media Broadband. So, if you're an existing Pay Monthly customer of O2, you'll need to get broadband from Virgin Media. Or, if you're already a Virgin Media broadband customer, you'll also need to get a Pay Monthly mobile with O2.

https://www.o2.co.uk/virgin-media-o2
 
Last edited:
Soldato
Joined
29 Aug 2006
Posts
4,104
Location
In a world of my own
Just had a virgin media local community lead knocking on the door this morning to let us know that in 3-4 weeks time Virgin are going to be running Fibre to our estate. I'm beyond pleased - been stuck with crappy BT 25Mbps service for nearly four years now and really hoping I can get the 1Gig service from virgin.

:)
 

TNA

TNA

Caporegime
Joined
13 Mar 2008
Posts
27,172
Location
Greater London
Just had a virgin media local community lead knocking on the door this morning to let us know that in 3-4 weeks time Virgin are going to be running Fibre to our estate. I'm beyond pleased - been stuck with crappy BT 25Mbps service for nearly four years now and really hoping I can get the 1Gig service from virgin.

:)
Holy Moley :D
 
Soldato
Joined
29 Dec 2002
Posts
7,176
Does anyone know why VM have really odd download to upload ratios? Seems a bit off that Gig1 is only 52Mbps upload.

Asymmetric ratio's are a technical feature of most 'broadband' services and have been since 512k ADSL. In the case of cable it's part of the DOCSIS specification, though more recent versions provide for symmetrical services at multi-gigabit speeds. Until the last few years, in effect all cable broadband was via a single strand of copper rather than fibre and that bit of copper was never intended to carry broadband signals at all. Even the pure fibre areas still rely on RFoG for 'compatibility' reasons, pure IP will make it a much fairer fight between OR/alt-net providers and VM, but at present and for practical reasons VM are going to have to take baby steps and stick with RFoG for a while yet.

Just had a virgin media local community lead knocking on the door this morning to let us know that in 3-4 weeks time Virgin are going to be running Fibre to our estate. I'm beyond pleased - been stuck with crappy BT 25Mbps service for nearly four years now and really hoping I can get the 1Gig service from virgin.

:)

A friend has them outside his house running ducting last week, he decided to try and find out when it would be live... apparently upto 18 months :D Somehow I doubt that.
 
Last edited:
Associate
Joined
6 Jan 2006
Posts
474
Just had a virgin media local community lead knocking on the door this morning to let us know that in 3-4 weeks time Virgin are going to be running Fibre to our estate. I'm beyond pleased - been stuck with crappy BT 25Mbps service for nearly four years now and really hoping I can get the 1Gig service from virgin.

:)
Welcome to my world. We've been stuck on BT/Plusnet 32Mbps for around 10 years with no speed improvement available. Virgin fibre was recently put into the street. As soon as I'm out of contract (February), I'll be moving to Virgin. And yes, 1 Gig was available from day 1.
 
Soldato
Joined
29 Dec 2002
Posts
7,176
Is the ping jitter and latency any better on VM FTTP or is it poo because the last mile still uses DOCSIS?

It’s a combination of legacy issues, VM has been making moves to bring the whole estate into line with other global LG territories using a standardised approach/supply chain, the commitment to lay fibre to every subscriber by 2028 is a massive step towards this, but I suspect they’ll stick with RFoG as an interim step and it’ll potentially be years after that before they can really get away from having to support legacy CPE/standards.
 
Caporegime
Joined
18 Oct 2002
Posts
26,052
Suspect their overbuild will be XGS-PON and they will gradually shift off RFoG as services are upgraded and customers churn. Especially if they want to start doing wholesale.
 
Soldato
Joined
29 Dec 2002
Posts
7,176
Suspect their overbuild will be XGS-PON and they will gradually shift off RFoG as services are upgraded and customers churn. Especially if they want to start doing wholesale.

Basically, yes. I would imagine wholesale would be 100% IP rather than anything else, you've already got VoIP as a standard from SH3 onwards, the hub branding/firmware isn't a problem, but they could go simple modems and use the standard - for example - Sky Hub, and VM's own IPTV has now soft lunched. Sky has now launched Glass which means they can now do IPTV themselves rather than just using Now and the Roku joint interest (how handy if you're looking at anchor tenant status on alt-nets and VM).

Migrations won't be a street by street approach like the Liverpool forced migrations from analogue, you'd have to replace every single bit of CPE if you ditch RFoG, that means significant head end changes and they simply haven't got the back end in place to provide a transition of all pay TV easily at this stage, even if they wanted to. It's only the end of last month they launched an IPTV service and it'll take a few years for the contract side to catch up in terms of licenses (conceivably upto 5 years if IPTV/cloud DVR wasn't specifically included last time round). It's easier and less expensive to let attrition/churn do the work, keep RFoG going till everyone is on fibre and then look at moving to pure IP for everything and phase out the older gen CPE eventually dropping RF. You also have the ability to charge a nominal fee for upgrades/installs this way. It's not ideal, but it's the only way that makes sense with the existing network.
 
Associate
Joined
6 Nov 2021
Posts
43
Location
Bristol
For anyone else, I fixed the Unbound issue. Now I have dnsmsaq serving dhcp and local DNS resolution for hostnames, and Unbound doing DNS over TLS to my AdGuard Home adblocking server upstream:

50DwFIk.png

OpenWrt actually also has the AGH package available itself, but I already have the VPS running and I'd rather have DNS off prem in case of issues, as our mobile devices rely on it off-network too. It was fairly easy in the end.

I downloaded and left seeding an Ubuntu 21.04 torrent to test latency this morning, and my BQM didn't even flinch. My base and average latency is down compared to before, no yellow spikes at all, and my upload is responsive throughout. Cake really rocks. Just a tiny increase in average latency during the seeding, and back down again once I closed the torrent client (after ratio >2). I've cropped it because it's such a small increase during seeding you'd struggle to see it on a full graph. It's around 5am-7am (notice the extra latency in the blue band). Tiny! Before I'd have had wide blue and massive yellow spikes. I can't recommend cake and proper SQM highly enough - now I know why Dave gets so excited, it's transformative to the line, and no more hiccups in WiFi calling or whatever when someone is saturating the upload.


7rojqW4.png

That is a lovely result.

While I'm hanging out here, is it possible to nominate Simon Kelly for a knighthood? He's british, and he's the author of dnsmasq (he's mostly retired now), which is probably still the most used dns server on the planet. He's very modest about it, and yet with quadrillions of dns and dhcp requests served... deserves a bigger boat.

I'm still very interested in "before/after" tests, but your results are typical of cake or fq_codel, typically wonderful, and yet a tad misleading.

1) You are showing zero packet loss on this test. There probably is plenty of that, but only on the heavier torrenting flows, which you can show by pasting the output of:

tc -s qdisc show

here, if you like. I can then try to explain what those statistics mean.

2) fq_codel is still a better, lighter weight algorithm than cake is in multiple respects, however several things that cake does that are very important are:

A) it does per host/per flow fair queuing. With fq_codel, Bittorrent in particular - say you have set it to 98 flows instead of the default 5- vs two regular download to another box, will still grab 98% of the bandwidth (although the ledbat algorithm helps a bit here, and the default of 5 is gentle and effective enough for transparent use of torrent with fq_codel)

it helps a lot in this extreme case to have cake on the bottleneck router where nat is. The nat option is part of dangerous settings in openwrt, (because a few devices don't do it right), but if that's on, and you have one device doing those two downloads, and 98 simultaneous torrent flows, each box gets half the bandwidth.

We didn't actually put that in there for torrent specifically, it came from a paper about gaming DASH (netflix-like) traffic: https://dl.acm.org/doi/pdf/10.1145/3344381

A way to think about per host fq is that each host gets *up to* 1/(active hosts) amount of the bandwidth. The up to part is important - if you are using less than your fair share
of the bandwidth, you don't feel the other traffic hardly at all. You can be under a massive DOS attack (from one server) and not notice, either. And most hosts are not active all the time.

B) cake also does diffserv classification. IF you want to assign even less bandwidth to torrent, mark it with either diffserv codepoint CS1 (deprecated) , or LE,
and that will end up in the bulk "tin", with a minimum of 5% of the bandwidth and a max of 100%.

C) Because you are doing inbound shaping you are observing a small increase in delay. If the ISP was doing this work for you, only in extreme cases would you see any
delay at all for ping, even 100 pings, at this bandwidth. That said, it's hard to trust the ISPs to get it right.

D) It does frame compensation right for docsis and dsl, letting you get very close to the ISPs's rate without error and thus controls latency spikes that can be induced by packets that are variable in size. You do have to tell it to use those parameters however.
Thx so much for burning the midnight oil to provide such an excellent example, and I hope all of England follows suit, or at least, deploys docsis-pie.
 
Last edited:
Back
Top Bottom