In reply to a post by jchamier:

VM's problem stem from the fact the cable networks were built by around 25 different companies, to different technical standards. So these effects are regional. Its why it is important to put your approx location on any post/complaint about VM.

I'm on the Surrey/Hants area (I think on the Guildford head end) and whilst I'm not a gamer, the BQM isn't as clean as an Openreach user, I don't get issues. (Live BQM link in sig).

In central London I was once a VM customer, for quite a long while, probably nigh on 15 years...this was ex Cable London, then Telewest/Blueyonder, then finally migrating to VM when they rolled in NTL.

Much of the latency issues are for want of better expression *architecturally intrinsic* to VM e.g the physical access architecture of DOCSIS is still (despite all the modernisation tweaks) effectively like the Ethernet of yore or CSMA/CD...and with the burden of TCIP/IP over the top = round trip delay nightmare.

Interesting article posted about 18 months ago on ISPreview about the potential (quite drastic) latency improvements possible with DOSCIS 3.1 and above. But have they actually materialised...seems not.

https://www.ispreview.co.uk/index.php/2019/06/docsis...

They're only using DOCSIS 3.1 for the downstream which is why there hasn't been any latency improvements (this is also why they sell "gigabit" Internet that's stupidly asynchronous).

Reading the white paper from CableLabs, it would appear that D3.1 in itself would not really resolve the latency issues. They would need to implement the further changes embodied in the Low Latency DOCSIS (LLD) addition to the specification.

In reply to a post by Pheasant:

In central London I was once a VM customer, for quite a long while, probably nigh on 15 years...this was ex Cable London, then Telewest/Blueyonder, then finally migrating to VM when they rolled in NTL.

We were CableTel which renamed itself NTL when it bought a non-cable business. Then they went and bought C&W Consumer Co, and eventually merged with Telewest (whose cable internet division was Blueyonder). The whole mess was renamed VM when they bought Virgin Mobile, and they now have to rent the Virgin name from Branson's Virgin Enterprises. (I believe VE has a 5% share holding!).

Much of the latency issues are for want of better expression *architecturally intrinsic* to VM e.g the physical access architecture of DOCSIS is still (despite all the modernisation tweaks) effectively like the Ethernet of yore or CSMA/CD...and with the burden of TCIP/IP over the top = round trip delay nightmare.

DOCSIS over glass (RFoG) seems to avoid that problem, so I think its more down to the physical plant deployment than the DOCSIS protocol.

Interesting article posted about 18 months ago on ISPreview about the potential (quite drastic) latency improvements possible with DOSCIS 3.1 and above. But have they actually materialised...seems not.

3.1 isn't properly deployed yet, a few years away!

In reply to a post by jchamier:

We were CableTel which renamed itself NTL when it bought a non-cable business. Then they went and bought C&W Consumer Co, and eventually merged with Telewest (whose cable internet division was Blueyonder). The whole mess was renamed VM when they bought Virgin Mobile, and they now have to rent the Virgin name from Branson's Virgin Enterprises. (I believe VE has a 5% share holding!).

They could cut the [censored] Virgin moniker adrift, just rebrand themselves in their native Liberty Global - save a bunch of cash and cut the baggage associated with the "V" brand. Its not like its got particularly glowing rep. any more these days....

In reply to a post by jchamier:

DOCSIS over glass (RFoG) seems to avoid that problem, so I think its more down to the physical plant deployment than the DOCSIS protocol.

I don't know to be truthful. I'm certainly no expert on DOCSIS but if the access architecture and 'queuing' is presumably the same whether the basic underlying carrier is copper in a coaxial cable or some SM fibre then how will that be better if the link back to the CMTS is glass - how would it make an appreciable difference - unless your using more RfOG as backhaul and either segmenting the network or distributing the CMTS to reduce congestions?

In reply to a post by jchamier:

3.1 isn't properly deployed yet, a few years away!

The Table 1 on the CableLabs white paper suggests that the 99th percentile values for latency don't alter until LLD is specifically introduced over D3.1 - in essence to deal with the serious inherent latency you need to completely re-work the access protocol to remove queuing. Not just run D3.1 on glass.

You say plant, I say protocol 😂

In reply to a post by Pheasant:

They could cut the [censored] Virgin moniker adrift, just rebrand themselves in their native Liberty Global - save a bunch of cash and cut the baggage associated with the "V" brand. Its not like its got particularly glowing rep. any more these days....

Maybe O2 Media in full time?

I don't know to be truthful. I'm certainly no expert on DOCSIS but if the access architecture and 'queuing' is presumably the same whether the basic underlying carrier is copper in a coaxial cable or some SM fibre then how will that be better if the link back to the CMTS is glass - how would it make an appreciable difference - unless your using more RfOG as backhaul and either segmenting the network or distributing the CMTS to reduce congestions?

I'm going by the significantly better BQM's from people on RFoG infrastructure. I assume significantly less sharing!

The Table 1 on the CableLabs white paper suggests that the 99th percentile values for latency don't alter until LLD is specifically introduced over D3.1 - in essence to deal with the serious inherent latency you need to completely re-work the access protocol to remove queuing. Not just run D3.1 on glass.

You say plant, I say protocol 😂

I'm certainly not a fan, but it meets my needs today until Openreach or an AltNet turn up

Interesting reading for me.
My daughter had a similar latency issue since she switched to VM [but not as bad as the OP's], and after about a month of me being on their case, her BQM rarely shows latency above 40, her speed is 10 times better but the WiFi from the hub3 is still Pants. After 4 weeks, 3 engineer visits, and a LOT of emails, VM claimed it was network congestion, and sorted it two days later. Maybe she was just lucky, but her neighbours have said their VM is better as well.
Farnborough Area
Gru

Edited by Grumpy20 (Thu 25-Feb-21 14:44:35)

In reply to a post by Grumpy20:

Farnborough Area

I'm in Farnborough, Hampshire. I have the Hub in modem mode, and use my own WiFi router.

Edited by jchamier (Thu 25-Feb-21 19:38:26)

In reply to a post by ian72:

Unfortunately there are fairly regular posts here about latency on Virgin and some people that move from Virgin back to a much "slower" FTTC because of the latency. You can try talking to Virgin but my guess is little will happen (unless they already plan to change something in your area).

Yep, that's me

550/35 Mbit VMB connection with frequent outages (unacceptable when it costs £80/month!) higher latency, bufferbloat, poor peering, and Puma6 latency issues.

Am on Zen FTTC now, 75/18Mbit.

Latency overall is twice as good as before. e.g. Google Stadia went from 20ms to 10ms
Transatlantic pings have also cut 10ms and the routes might be a bit faster.

Though I really miss the fast download speeds.

https://www.youtube.com/watch?v=GlRQjzltaMQ

Been this way for years.