120M : Can someone explain why cable suffers on upload?

[Risco]

I know it is something to do with how it works and is the reason why the upload speeds are low and need capping to stop over utilisation. That is about it though!

Wondering how it differs from fibre, and why fibre does not suffer the same issues. Am I right in saying that is a problem that can't be solved?

[Sephiroth]

The upstream up to recently operated only between 18 MHz and 45 MHz frequency. That allows about 4 upstream channels at 6.4 MHz bandwidth. The range used by VM has now expanded to 65 MHz which puts 8 channels available at that bandwidth.

This compares with the downstream where each node if you're lucky and each segment in less well structured areas, has 16 downstream channels at a frequence range 10 times higher.

The upstream frequency range is somewhat more sensitive to RF noise than the downstream. So VM can't put the bit density up as high as on the downstream. VM are moving upstream from 16QAM to 64QAM, expanding capacity per packet by 50%. By comparison downstream currently operates at 256QAM (an extra 33% packing density).

You mention fibre. What do you mean by that? BT FTTC (Infinity)? If so, their frequency plan is totally different allowing a faster upstream to be offered.

[kwikbreaks]

Plus a downstream channel has 50Mbps capacity but upstream it's only 18Mbps.

On the downstream side there is no "request to send" necessary and the CMTS can just about completely fill the available capacity while each modem listens to everything all the time and picks out any traffic directed to it. On upstream you can't have every modem transmitting at the same time or they would interfere so there is some overhead organising when each can transmit. That's probably the primary reason why cable suffers so badly from jitter as with xDSL a modem can transmit anytime it needs to because it is the only modem on that particular circuit.

[Sephiroth]

Yes - Kwikkie's filled the gap. An upstream packet (for want of a better term) has to request a fixed size "mini-slot" on each of the two bonded channels. On a busy upstream (p2p is on spoiler; the Cloud is another), then games have to wait for that slot on the basis of queing theory.

[qasdfdsaq]

Originally Posted by Sephiroth

You mention fibre. What do you mean by that? BT FTTC (Infinity)? If so, their frequency plan is totally different allowing a faster upstream to be offered.

I've told you several times before but this is entirely misleading. The "frequency plan" of the two systems cannot be compared in this way, and also the frequency plan has nothing to do with why the two systems offer differing upload speeds and capacities. The term "frequency plan" doesn't even apply in the same way to either technology.

By your reckoning,
VM upstream operates between 18 Mhz and 45 Mhz and contains (up to) 4 upstreams of 6.4 Mhz each

BT upstream operates between 0 Mhz and 17 Mhz and contains two channels of 2.4 Mhz (actually one 1.6 and one 3.2)

How on earth does the latter allow a faster upstream to be offered?

---------- Post added at 11:29 ---------- Previous post was at 11:21 ----------

Originally Posted by Risco

I know it is something to do with how it works and is the reason why the upload speeds are low and need capping to stop over utilisation. That is about it though!

Wondering how it differs from fibre, and why fibre does not suffer the same issues.

The real answer is very simple and it boils down to the fundamentals of how the two technologies work.

On Virgin Media cable, your cable is shared with dozens to hundreds of other people. Each cable may have enough capacity for 3 connections' worth and 300 people trying to use it. Virgin Media are counting on you not trying to all use it at the same time.

On BT/Openreach fibre, every single cable is dedicated and provides individual bandwidth to every single user. You get the full capacity of your line all to yourself. You're not sharing your speed between 300 other users.

Am I right in saying that is a problem that can't be solved?

Not entirely. VM can break the cable bundles (nodes) down so they are shared between fewer and fewer people, and this is an ongoing effort that all cable companies are trying to achieve - but each step costs exponentially more (although still less than it's costing BT to roll out their completely new fibre network)

[Risco]

Originally Posted by qasdfdsaq

I've told you several times before but this is entirely misleading. The "frequency plan" of the two systems cannot be compared in this way, and also the frequency plan has nothing to do with why the two systems offer differing upload speeds and capacities. The term "frequency plan" doesn't even apply in the same way to either technology.

By your reckoning,
VM upstream operates between 18 Mhz and 45 Mhz and contains (up to) 4 upstreams of 6.4 Mhz each

BT upstream operates between 0 Mhz and 17 Mhz and contains two channels of 2.4 Mhz (actually one 1.6 and one 3.2)

How on earth does the latter allow a faster upstream to be offered?

---------- Post added at 11:29 ---------- Previous post was at 11:21 ----------


The real answer is very simple and it boils down to the fundamentals of how the two technologies work.

On Virgin Media cable, your cable is shared with dozens to hundreds of other people. Each cable may have enough capacity for 3 connections' worth and 300 people trying to use it. Virgin Media are counting on you not trying to all use it at the same time.

On BT/Openreach fibre, every single cable is dedicated and provides individual bandwidth to every single user. You get the full capacity of your line all to yourself. You're not sharing your speed between 300 other users.


Not entirely. VM can break the cable bundles (nodes) down so they are shared between fewer and fewer people, and this is an ongoing effort that all cable companies are trying to achieve - but each step costs exponentially more (although still less than it's costing BT to roll out their completely new fibre network)

Ahh now that makes sense. So essentially they can relieve the problem by investing in the network but cannot eradicate. I always thought Virgin was fibre optic, so why can they not offer the same dedicate line?

[kwikbreaks]

The last leg on VM's network is a shared coax cable which serves many subscribers.

The last leg on xDSL is a twisted pair to each individual subscriber.

The real difference between the two is that with VM a low bandwidth pipe (current max 400Mbps down 36Mbps up) is shared by a relatively small number of subscribers and with xDSL a much larger pipe gets shared by a much larger number of subscribers and you are less likely to see congestion problems with that setup.

[qasdfdsaq]

I don't think that's a fair comparison, because on VM you also have a large pipe shared between a much larger number of subscribers at the UBR and core. The whole point is VM has an additional shared last mile whereas xDSL does not. This moves the bottleneck and contention point to another part of the network that doesn't exist in the competing technology.

There's no significant difference between the "large pipe" mechanics at the UBR and at the DSLAM. The difference is caused by the section between those two and the end user - one is shared, the other is not. Upstream of that everything is shared on any ISP and no technology has any advantage over any other.

[kwikbreaks]

Yes I forgotten to specify that the comment on pipe size was specific the the final section of the network. I'd assumed that was obvious.

[Dark Fiber]

I think the OP is asking "why is cable asymmetric?" The answer is surely because it started out as a television distribution system with little need for a back-channel and has subsequently been bodged to carry internet. Most if not all of the problems listed by expert commenters above could have been designed out if the growth of home internet had been anticipated.

[kwikbreaks]

I think all home broadband is asymetric because there is a limited total bandwidth that cable and twisted pair can carry and favouring downstream suits the majority of users. For most protocols 5% upstream (eg for acks) is ample to keep the downstream running at full tilt.

[General Maximus]

Originally Posted by Risco

I always thought Virgin was fibre optic

it is another one of those marketing gimics they seem to get away with. Only part of the network runs on fiber and I believe that is from an active cabinet back to the UBR and beyond. In contrast, open reach is fiber all the way to your street cabinet which makes their assertion less of a lie.

[Kushan]

I think BT's fibre is now as much fibre as Virgin's, i.e. hybrid. The difference ultimately being that Virgin's last mile is shielded coax, whereas BT's is unshielded (or less-shielded, I suppose) copper. The technologies in use, of course, are completely different.

[Sephiroth]

Originally Posted by Kushan

I think BT's fibre is now as much fibre as Virgin's, i.e. hybrid. The difference ultimately being that Virgin's last mile is shielded coax, whereas BT's is unshielded (or less-shielded, I suppose) copper. The technologies in use, of course, are completely different.

I think, Kush, that's on over-simplification. Fibre is semantically fibre. It's what's done with it that's the difference.

The FTTC fibre node has a strict limit (AFAIK) of 288 users. A VM optical node might well have twice as many.

[Kushan]

As I said, the technology is completely different. What I was referring to was how much fibre it takes before you can say you've got a fibre network. Technically even the ADSL lines eventually connect to a fibre backbone but that doesn't count.