Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[meritez]

plenty of feedback on the virgin forum about 1gig service.

one should remember the hub they supply only has gigabit ports, so once you factor in the actual real-life achievable speed form a gigabit port you'll end up with around 940Mbps being supplied by the hub (even though the actual speed being supplied to the hub is usually about 1.2Gbps).

If you know what you are doing, it's possible to get more than 940Mbps out of the hub 4:
https://tech.msh100.uk/virgin/networking/2020/10/17/virgin-media-greater-than-gbit/

[Chrysalis]

It's not true it's 108 Mbit between everyone on the CMTS, no.

A high tier user won't consume 20% of it downloading as cable has mechanisms to suppress and accumulate acknowledgements.

Localised congestion is still a thing and might be for a little while but should all be sorted in time through 3.1 upstream and Project Mustang.

Hmm thats good and bad, not sure I want my acks to be manipulated by the isp.  How much shared upstream is there if you dont mind answering? For some reason I cant find this info, I remember a decade ago this stuff was much easier to find.

[Alex Atkin UK]

If you know what you are doing, it's possible to get more than 940Mbps out of the hub 4:
https://tech.msh100.uk/virgin/networking/2020/10/17/virgin-media-greater-than-gbit/

Seems a silly configuration, they even admitted as much in the comments.  Why waste the the second switch when two ports will do the job just fine leaving the other switch able to be used.

Fair enough if they want to do it with four, but when sharing the idea publicly its unhelpful to overcomplicate things.

[Reformed]

Hmm thats good and bad, not sure I want my acks to be manipulated by the isp.  How much shared upstream is there if you dont mind answering? For some reason I cant find this info, I remember a decade ago this stuff was much easier to find.

Each CMTS has multiple 10G uplinks. Each cable segment has 4-6 x 27 Mbit/s upstream channels. A single Gig1 user has half the capacity their modem is addressing. VM have major problems with network noise due to regulation preventing them taking stronger measures to remove it.

The acknowledgement suppression has been a thing for 20-odd years. It doesn't break anything.

EDIT: It's also not done by the ISP, cable modems do it. Rather than sending every acknowledgement separately if they have acknowledgements arrive while waiting for upstream slots they'll acknowledge everything cumulatively. Saves upstream capacity all around.

Also on the matter of the guide I'm going to cut the guy all the slack in the world given he gave us https://bidb.uk/

[Chrysalis]

Ok so the acks are delayed for a few ms so they just jittered, so still same bandwidth cost just better allocation of timeslots.  Your wording made it sound worse.  The shared upstream segment is between 108 and 162mbps.  Thanks.

A guy who claims to work for VM has told me there is no giga1 customers for every cabinet on my street meaning I would be the only one on the docsis 3.1 DS channel, and the average speed for the one 500mbps customer is basically full throughput both directions.  Although I gave him my postcode he then wanted my full details which I have refrained from as there is no way of knowing if he is legit.

[Reformed]

They aren't jittered, it cumulatively acknowledges each TCP stream at each slot rather than buffering them and sending them all. Cable modems have a bunch of functionality above layer 1 and 2.

[Alex Atkin UK]

Also on the matter of the guide I'm going to cut the guy all the slack in the world given he gave us https://bidb.uk/

Fair point, and I can see how he might not have wanted to document a slightly different method he didn't test himself, even if he "thinks" it should work fine.  I do forget sometimes that not everyone has a lot of free time for this stuff.

Now if he could just fix that site so the coverage dots work when zoomed in.  He DOES seem to have finally fixed the weird scaling glitch I was having which the key would keep jumping between two different sizes.

[Chrysalis]

They aren't jittered, it cumulatively acknowledges each TCP stream at each slot rather than buffering them and sending them all. Cable modems have a bunch of functionality above layer 1 and 2.

Jitter was perhaps the wrong word to use, but cumulatively adding is still sending them all, so I am misunderstanding you and will leave it here.

Depending on what cityfibre come back with, I will order VM, they have a cooling off period, so I can analyse actual ACK traffic levels to answer my own question then, and see if it works ok during peak for what I want it for.

[Reformed]

You won't see anything on the cable modem side. Your router will send as normal. You can of course capture on a remote server if that turns you on.

[Chrysalis]

Yep I was going to check server side if ACK's were been modified to remove data to make them smaller or just batched together.

[Reformed]

Okay I'm not sure if I'm being clear.

Cumulative acknowledgement is a specific technical term and I was using it that way. It means to send a single acknowledgement to acknowledge all of a stream received so far.

Cable modems behave as layer 4 proxies and send a single acknowledgement covering all the acknowledgements they have received for a particular stream.

You will see far fewer acknowledgements at the server-side than your client machine is transmitting. This neither breaks or degrades anything as loss is handled by selective acknowledgement.

[Chrysalis]

Ok does anything break if selective acknowledgements are disabled?  You are right that after you gave me the technical term I understand what you was saying better.

Thank you for providing it as it is interesting to learn something new.

[Alex Atkin UK]

That's very useful as I've seen instructions for tweaking NACK and SACK but never understood what they were.

[Reformed]

Ok does anything break if selective acknowledgements are disabled?

Not that I'm aware of, Sir.

[Alex Atkin UK]

Not sure why you'd want to.

TCP Selective Acknowledgment (SACK) is used to improve performance of data transfer on TCP stack.

Disabling TCP SACK will negatively effect the data transfer performance. The magnitude of performance hit depends on type of data transfer.

Its basically mitigating the delay of waiting for an upload time slot by reducing how many ACKs need to be sent when that time slot becomes available.