In reply to a post by New_Londoner:

Good to know that cable upstream speeds might finally be getting competitive, however does anyone know how many broadband users at 100Mb the shared coax segment that can link 1000+ properties is capable of supporting?

In other words, are cable customers going to suffer from increasing contention on the coax at busy times of day, with throughput dropping off due to lack of coax bandwidth rather than any problems with backhaul etc? Is the architecture up to the task as Virgin introduces higher bandwidth options?

OK.

A few things here.

Cable bandwidth should be considered in terms of kbps per modem, rather than any absolutes, that's the key figure.

The architecture is fluid, the main way of upgrading an HFC network is node splits, reducing the amount of people sharing a single port / card through the use of existing or new fibre to supply dedicated links back to the hubsite or headend.

The 100Mbit/s and 10:1 upstream ratio involves a standard upgrade cycle of the following:

Splitting a node, then splitting each split, quadrupling the downstream bandwidth per home.

Following the split upgrading the upstream channel width, further doubling the upstream bandwidth per home for a total 8-fold increase.

Noteworthy also is that while there are some areas left which are 1000 homes passed, however it should be noted that most now are 250 - 500 homes passed nodes, uptake of cable broadband will be around the 40 - 50% mark, and of that 400 - 500 modems connected the majority will be at 10Mbit.

Virgin will be supplying either 1 or 2 38Mbit channels which take care of the 10Mbit customers and a further 4 50Mbit channels which look after both the 20 and 50Mbit customers from this node and potentially another 3 nodes, and will do so with no real issues.

This is quite non-trivial to explain without understanding the basics, fundamentally no, the existing architecture isn't up to the task which is why they are upgrading it by pushing fibre deeper and through this reducing the number of homes sharing the coaxial trunks an in turn the fibre.

The 'how many users' question is a piece of string question. The answer is based on such things as:

Usage patterns
QoS in use
Application level management, if any
What level of contention is tolerable, what is the target minimum throughput

etc.

I've seen such huge swings - I've seen 38Mbit/s happily serving 1200+ modems, when the maximum speed available was 2Mbit/s, and likewise seen 38Mbit/s caned by 50 modems - it's impossible to give even a ballpark figure.

if it helps my area runs just fine, perfect speeds 24x7 and there are well over 300 20Mbit and 50Mbit modems connected to the downstream port. There are 4 upstream ports serving the downstreams giving a total bandwidth of 200Mbit/s downstream and 4 x 9Mbit/s upstream, each modem has approximately 0.6Mbit/s of peak time bandwidth which is more than adequate, the port doesn't even reach 50%, or 100Mbit - people's usage is, still, generally quite intermittent or relatively low bandwidth.

Following the upgrade cycle I mentioned above each modem would have 2.4Mbit/s of downstream bandwidth, which would, worst case if every single modem is on 100Mbit which simply isn't going to happen, give a worst case contention of a shade over 40:1. Given the split will be heavily biased towards 20Mbit/s modems the actual ratio is likely to be more like 15:1 - 20:1 - ample under virtually any circumstances.

In reply to a post by New_Londoner:

Thanks.

So can a shared coax segment cope with every customer taking 100Mb simultaneously? Obviously worst case example, interested whether there is enough bandwidth available for broadband on the coax to allocate sufficient channels to every user, assuming 1000+ homes connected?

No residential user network can cope with that, full stop.

Per my above comment also cable upgrades generally aren't allocation of new channels but are reduction of numbers of modems sharing channels.

In reply to a post by utis:

Any details on when to expect 100mbit UK wide release? What upstream options will be? What pricing? Will 50mbit users be automaticallly upgraded to 100mbit? Thanks..

Middle of next year for UK wide completion, will be a staggered rollout same as the 50 was, starting before the end of this year.

10Mbit will be the upstream.

No idea on pricing.

50Mbit will not be upgraded to 100Mbit, you will, however, get a 1.75Mbit -> 5Mbit upstream upgrade as part of the all tier upstream upgrade to 10:1 ratio.

In reply to a post by ElBobbo:

[Edit: Re-reading what you said, it's my understanding that the trunk coaxial cables can be saturated as each drop along the distribution line does not have a discrete/exclusive signal range. I would imagine that the fibre backhaul is much more heavily contended than the coax paths and thus will cause problems much sooner.]

The fibre back from the cabinets is a simple media conversion from coaxial to optical. For broadband purposes congestion on cable networks should only be considered in the context of CMTS port utilisation.

In reply to a post by Ignitionnet:

The fibre back from the cabinets is a simple media conversion from coaxial to optical. For broadband purposes congestion on cable networks should only be considered in the context of CMTS port utilisation.

You're much more likely to reach saturation on the backhaul feeding the CMTS (channels are cheap, backhaul is generally not).

In reply to a post by ElBobbo:

In reply to a post by Ignitionnet:

The fibre back from the cabinets is a simple media conversion from coaxial to optical. For broadband purposes congestion on cable networks should only be considered in the context of CMTS port utilisation.

You're much more likely to reach saturation on the backhaul feeding the CMTS (channels are cheap, backhaul is generally not).

On what information do you base this?

Backhaul along one's own fibre network, per Virgin Media and pretty much every other cable company of any size, they do after all own physical networks, is cheap. CMTS cards and node splits potentially involving new fibre pulls to cabinets and new cabinet construction not so much.

Channels are most certainly not cheap, each downstream group needs a corresponding upstream, upstream bandwidth is expensive on cable networks and in areas where the downstream network is busy they're likewise expensive.

If it were as simple as just piling channels in Virgin wouldn't split nodes, it only delays the inevitable. Telewest tried that approach while waiting to be bought by ntl which is why Virgin have had to spend so much money sorting out ex-Telewest areas through belated node splits.

Some ex-Telewest areas, this being one of them, have a maximum of 25MHz of upstream bandwidth available, with about 12.8MHz of that actually usable simultaneously due to network conditions and again due to network conditions this totals out to about 36Mbit/s of upstream bandwidth being available to the area. Noting that DOCSIS upstreams begin to degrade at about 65% utilisation and totally suck by 80% or so you're talking planning around say 25Mbit/s of upstream utilisation, which is split across 4 channels, so you're talking 4 channels of just over 6Mbit/s each which can be used to support downstream traffic. Not a hell of a lot of point in piling 4 x 50Mbit/s bonded groups onto each upstream channel to relieve downstream issues, upstream will soon become an 'issue'.

Removing this restriction on the upstream bandwidth, and where networks are build as such downstream as well, requires new optics and new amplifiers in the field for each and every fibre optic node, trunk, bridge and line that needs the upgrades. Given each amplifier can be serving just a handful of homes and each node likely sub-500 customers it gets kinda expensive kinda quickly.

It's a tad more expensive per home than a new 10GbE PIC on the T640 and a new 10GbE XENPAC on the local switch / GSR, which will serve for the primary path for a large town's hubsite.

In reply to a post by Ignitionnet:

On what information do you base this?

Outdated information, it seems.

Thanks for the clarification.

Is cable up to the task?

http://optimum.custhelp.com/cgi-bin/optimum.cfg/php/...
http://www.upc.nl/internet/
http://www.upc.pl/internet/
http://www.comhem.se/
http://www.zon.pt/Internet/TarifariosFibra/Detalhe.aspx (The 1Gbps is FTTH)

etc

Seems to work ok elsewhere

In reply to a post by ElBobbo:

In reply to a post by Ignitionnet:

On what information do you base this?

Outdated information, it seems.

Thanks for the clarification.

Was never the case, whoever gave you that info was pretty misinformed. Backhaul hasn't been an 'issue' on either network ever really. The exception being the occasional fault with the MetNets.

Peering on the other hand....

Why do you always want more?