In reply to a post by Ignitionnet:

Openreach's allergy to CapEx has become abundantly clear, and areas where they're deploying a second FTTC cabinet at considerable cost, and will follow it up by deploying G.Fast, and follow that up by eventually deploying FTTP are a testament to that attitude. They're like someone running on a payday loan who'll pay way more in the future to avoid having to pay now.

Brilliant

Is like I said about the ECI, cheaper for openreach to have got the vectoring capable hardware of the bat instead of getting the cheaper stuff first and then a few years later replacing it, more long term cost but with the gains of saving money in the immediate short term, your payday description fits what they doing perfectly

BT seem to love milton keynes, they have had a lot of special treatment in the last half a dozen years or so. I wonder which BT exec lives there.

I don't doubt that BT have made some strategic errors here and there, but I really don't buy the theory that Openreach is going to find itself running FTTP to every residence any time soon, nor that we're going to be kicking ourselves as a nation that they don't.

Some of you might have seen the coverage of Netflix's arrangement with the big ISPs in the US. There's talk of how video is very demanding, and the costs involved... but if you looked very carefully you'd find some hard numbers. Netflix customers typically run at around 2-3Mbps. That's on FTTP and FTTN products, and it's _after_ Netflix spent a bunch of money against their will to arrange direct peering. That's "Now our customers are happy with how speedy everything is" bandwidth in other words. But it's a number you could achieve with ADSL2+ if only everyone lived near enough to an exchange. My point is that one of the most "bandwidth intensive" popular applications uses so little.

The reason BT's Infinity adverts focus on a multi-user household with intensive needs is that aside from bragging rights it rapidly gets harder to justify more bandwidth as the numbers go up. FTTC puts a huge number of residences beyond the point where "bandwidth" is something they seriously need (and so will pay for) more of, even on the upstream side. So it's just never going to make sense to provide those residences with FTTP. For other issues like latency due to "buffer bloat" more bandwidth is an expensive sticking plaster, not a real solution.

People tend to wrongly project from a few data points on an upward trend that things will grow endlessly, but in reality they're more often on something like an S curve, and they just can't see that yet.

I bet early electricity distribution looked like this too. At first you're running a cable to someone's home for a light, then it's five lights, then they want to run a refrigerator, oven, heating... it seems to analysts as though demand grows without end, but they're wrong. Running 1MW electrical supply to each home would not be "future-proofing" it would just be wasted money.

In reply to a post by tialaramex:

I don't doubt that BT have made some strategic errors here and there, but I really don't buy the theory that Openreach is going to find itself running FTTP to every residence any time soon, nor that we're going to be kicking ourselves as a nation that they don't.

The percentages on this news story are concerning.

The majority of those not covered by Virgin cable seeing 38Mb/s and below. It's debatable how future proof that kind of performance is.

The long term value of deploying 1 or 2 FTTC cabinets to an area, along with the operating costs of powering them, line cards, etc, then deploying G.Fast to distribution points, then eventually deploying FTTP compared with deploying FTTP sooner is debatable.

For other issues like latency due to "buffer bloat" more bandwidth is an expensive sticking plaster, not a real solution.

Bit confused. What else is to be done about links that are saturated, hence having to use excessive buffering to try and manage traffic, beyond upgrade their bandwidth? Do we just strip buffers and ditch packets as soon as they come onto the line? How then do ISPs manage QoS on the links, nailing up bandwidth for businesses while running best effort residential servics?

Edited by deleted (Sat 03-May-14 15:10:02)

ignitionet - as you are will aware there is a challenge between the BAU work and Fibre programme -you comment around he The 2.5 billion commercial expenditure figure that Openreach bandy about is likewise laughable -is unhelpful as you know that to be fact aas you have heard that from a number of senior people your cab was finailly covered under that banner -- as you know developers tend to spend money on new community centres and Schools as part of their 106 - dont engage the infrastrcure provdier about what they are really buildin ans then wonder why the broadband is not as they expect to to be

FYI less FTTP equals more FTTC and more exchanges have become commercial (under the 2.5bn) (which were originally non commercial due to the increase in FTTC and Decrease in FTTP the intervenion areas on some of the later counties are smaller than they could have been due to th fact you can do more FTTC (exchanges) than you can by buildind FTTP

so for avoidance of doubt it is £2.5bn and it was not public money -so the business makes it decisions on this - to the beneift of shareholders / investers
if a commnity want FTTP the they can look at funding it as you are well aware
FYI whats happening with cab 90 ?

Edited by deleted (Sat 03-May-14 17:05:46)

In reply to a post by Fastman2:

ignitionet - as you are will aware there is a challenge between the BAU work and Fibre programme -you comment around he The 2.5 billion commercial expenditure figure that Openreach bandy about is likewise laughable -is unhelpful as you know that to be fact aas you have heard that from a number of senior people your cab was finailly covered under that banner -- as you know developers tend to spend money on new community centres and Schools as part of their 106 - dont engage the infrastrcure provdier about what they are really buildin ans then wonder why the broadband is not as they expect to to be

FYI less FTTP equals more FTTC and more exchanges have become commercial (under the 2.5bn) (which were originally non commercial due to the increase in FTTC and Decrease in FTTP the intervenion areas on some of the later counties are smaller than they could have been due to th fact you can do more FTTC (exchanges) than you can by buildind FTTP

so for avoidance of doubt it is £2.5bn and it was not public money -so the business makes it decisions on this - to the beneift of shareholders / investers
if a commnity want FTTP the they can look at funding it as you are well aware
FYI whats happening with cab 90 ?

The 2.5 billion figure is CapEx and OpEx. The impression is given all over that that figure is for network build which isn't really the case. The last update to investors indicated incremental CapEx of 1.3 billion.

Senior people claim an exchange costs a million and then each cabinet 100k on top to enable for FTTC. Those words came from the then-Openreach CEO and are not the case. Even the final 50 million in additional commercial funding is only costing 125GBP / premises passed.

I believe originally 25% of the 19 million premises roll out was planned to be FTTP. Given the cost differences between the two I would've hoped that dropping from 4.75 million FTTP premises planned to not even 250,000 funded solely by Openreach without ERDF / council funding would've meant quite a lot more FTTC than originally planned.

Regarding self-funding, I was actually wanting to buy FTTPoD sadly I'm in band F, and the exchange was enabled just in time for the price increases to kick in and prevent my ordering. The 100k for this cabinet appears as if it didn't cover bringing the fibre spine closer than a kilometre away.

The install charge went from 500+VAT to 750+VAT, the distance component from 2.5k+VAT to 4.375k+VAT, which still wasn't enough to scare me, and the annual fee from 456+VAT / year to 1188+VAT / year before it gets to the exchange, which does.

Cabinet 90 was enabled before 82, as part of the first infill of Hunslet exchange. I presume you're referring to cabinet 91.

The estates are hybrids of MDUs, terraced, semi-detached and detached housing so a few options are possible. I'm in touch with one supplier over possible hybrid build of FTTB to the flats and FTTP to homes in that area and, in turn, over a wider area of this estate. Any future self-funded deployment will almost certainly be to an alternative supplier.

I'm aware the business decision is to the benefit of the investors - in the short term. They're the ones holding the purse strings. In the case of a cabinet like this one in the longer term Openreach face potentially the costs of standing another cabinet, line cards, and powering them both. Then at some point later on the costs of potentially building a new aggregation node, the current one is a way away and unsure how much capacity it has, but most certainly costs of deploying G.Fast to distribution points. Finally at some point in the future the cost of completing the fibre drops.

Of course this assumes that Openreach won't be tapping up the public sector for money again.

Meanwhile the people on Virgin Media are getting higher downstream speeds than I am, even with 2 x FTTC lines bonded, for less than half the price I pay for the two lines alongside their line rental. The next thing I have in prospect to start to catch up is vectoring which seems to have made very little progress. Being in an Openreach-only area I have no other choices other than to pay over 200GBP/month retail and 6k up front for FTTPoD.

The choice that BT are so concerned about providing to end users seems to only extend to the very low end, the ability to be able to pay next to nothing for second-rate DSL from A N Other cheap supplier. Wanting a contended, best effort product that's between EAD and xDSL pricing seems to be unreasonable though.

Obviously Ofcom are partly to blame, but there's been very little sign in public that Openreach are trying to change things. The 'Fibre Only eXchange', Deddington, isn't fibre only at all, the copper is all still there. Why aren't Openreach out in public trying to get this changed, it's crazy? FTTP would make more commercial sense overnight if Openreach were permitted to remove copper, but I guess it is too profitable to want to do so.

In reply to a post by Ignitionnet:

The majority of those not covered by Virgin cable seeing 38Mb/s and below. It's debatable how future proof that kind of performance is.

Sure, and I summarised the debate earlier. Without as yet unimagined novel high bandwidth applications 20Mbit/s feels very nice even with several simultaneous users, subject to the caveat that I'll get to below.

... then eventually deploying FTTP compared with deploying FTTP sooner is debatable.

Sure, if you're determined to imagine that the trend continues to infinity, which is exactly the scenario I was arguing against.

Bit confused. What else is to be done about links that are saturated, hence having to use excessive buffering to try and manage traffic, beyond upgrade their bandwidth? Do we just strip buffers and ditch packets as soon as they come onto the line? How then do ISPs manage QoS on the links, nailing up bandwidth for businesses while running best effort residential servics?

See? This sounds intuitively right doesn't it. And that's been good enough for people designing things like your average home WiFI AP, DSL modems and so on. It isn't true though, as experienced network engineers knew, but nobody thought to ask them. Of course once they realised why their home network performance sucked so badly they began working on the problem, but first let's see why it's a problem at all and why the intuitive understanding is wrong.

Imagine we have 1Mbps network link, and our packet size is up to 1500 bytes. After we start sending a packet we have to wait 12ms before we can choose what to send next. Even if we realise the very moment after we start sending that we wished we were sending a different packet, it's too late. Still it's possible to achieve acceptable VoIP over this network, I hope you agree, so long as we choose any outstanding VoIP packet to go next if there is one.

Now, suppose we add a small buffer to the hardware. The buffer is 9 kilobytes, enough for six of those 1500 byte packets. The hardware accepts packets right away as long as it has space in its buffer, and they are sent on a First In First Out basis. When the network is lightly used, the buffer will be empty, and our VoIP packets behave much as before, with 12ms latency imposed by the slowness of the network. But when the network is busy, the buffer fills up, and newly sent VoIP packets find themselves behind not one but six other packets, latency climbs to 72ms. The "good" news is that in a straight forward bandwidth test the buffer-equipped hardware comes off slightly better. A reviewer may find that they see 0.98Mbps reported on a test whereas with the "no buffer" hardware it was only 0.96Mbps. So it's better, right?

Well if 9 kilobytes of buffer RAM helped, you can be sure the manufacturer will try adding more. Here's a new model with 64kB of buffer. Reviewers report that now the bandwidth measured was shown as 0.99Mbps, very impressive. They don't measure the VoIP latency when the network is busy, but we shall, as we'll find that it's now over 500ms. Conversations are frequently confused, people interrupt each other, and soon they're saying "Don't use the Internet, I'm trying to make a call". How unsatisfactory.

That's Buffer Bloat, you can read more about it by searching for the term. Most buffers in network hardware are superfluous, added by people who didn't understand the problem they were trying to solve. Because so often people focus on bandwidth to the exclusion of latency they don't realise they've made things worse.

igntiionnet- there is no case for widescale FTTP (its just hard and complcated and expensve and indivually bespoke - it only really works in newbuild when tou can deployit as a greenfiels site) - i sure you are aware of the payback on commercial network -- and BDUk is even longer -- their is no case for FTTP - openreach still have to provide voice serivices as part of its operating licence so lots of FTTP and remove the opper sounds great in tehory - but openreach has to operate ithe real world and it has to be proven to be robust (so reality is another challenge all together) -- can you image the furore of your house buring down and not being able to place a emergency call could not be placed due to voice network not operating -- (copper provides Voice) and wil continue to do so for a considerable period of time (i agree that FTTP is a preference in in new sites but then if you dont do any copper you have to deploy FVA Voice and there is only one provdier deploying that at present - that is also dependant on the develeoper just deploying copper as you know to your cost

Edited by deleted (Sat 03-May-14 21:31:12)

In reply to a post by tialaramex:

See? This sounds intuitively right doesn't it. And that's been good enough for people designing things like your average home WiFI AP, DSL modems and so on. It isn't true though, as experienced network engineers knew, but nobody thought to ask them. Of course once they realised why their home network performance sucked so badly they began working on the problem, but first let's see why it's a problem at all and why the intuitive understanding is wrong.

Imagine we have 1Mbps network link, and our packet size is up to 1500 bytes. After we start sending a packet we have to wait 12ms before we can choose what to send next. Even if we realise the very moment after we start sending that we wished we were sending a different packet, it's too late. Still it's possible to achieve acceptable VoIP over this network, I hope you agree, so long as we choose any outstanding VoIP packet to go next if there is one.

Now, suppose we add a small buffer to the hardware. The buffer is 9 kilobytes, enough for six of those 1500 byte packets. The hardware accepts packets right away as long as it has space in its buffer, and they are sent on a First In First Out basis. When the network is lightly used, the buffer will be empty, and our VoIP packets behave much as before, with 12ms latency imposed by the slowness of the network. But when the network is busy, the buffer fills up, and newly sent VoIP packets find themselves behind not one but six other packets, latency climbs to 72ms. The "good" news is that in a straight forward bandwidth test the buffer-equipped hardware comes off slightly better. A reviewer may find that they see 0.98Mbps reported on a test whereas with the "no buffer" hardware it was only 0.96Mbps. So it's better, right?

Well if 9 kilobytes of buffer RAM helped, you can be sure the manufacturer will try adding more. Here's a new model with 64kB of buffer. Reviewers report that now the bandwidth measured was shown as 0.99Mbps, very impressive. They don't measure the VoIP latency when the network is busy, but we shall, as we'll find that it's now over 500ms. Conversations are frequently confused, people interrupt each other, and soon they're saying "Don't use the Internet, I'm trying to make a call". How unsatisfactory.

That's Buffer Bloat, you can read more about it by searching for the term. Most buffers in network hardware are superfluous, added by people who didn't understand the problem they were trying to solve. Because so often people focus on bandwidth to the exclusion of latency they don't realise they've made things worse.

Sorry, I'm just a network performance guy for a living, and I still don't get your point. Without buffering an overbooked network will simply replace latency with packet loss, and without queuing, which requires buffers, it's impossible to apply QoS to a network, which is essential in enterprise environments.

Surely having enough bandwidth out of the buffer to ensure it doesn't start to fill is the way to go, or if that's not feasible the approach that I've seen more commonly which is not to use FIFO but to use slightly smarter queuing which favours smaller packets over larger ones?

You're aware of weighted fair queuing, for example, which works well to avoid single high bandwidth flows hogging bandwidth, or class-based H-FSC which is effective as a scheduling algorithm across TCP and UDP traffic?

Experienced network engineers tend to see their own bit but tend to ignore what's actually running on the network at layer 4 and above. A VoIP call can cope with a little delay, what it can't cope with is loss. My employer's products are built both to apply QoS and to mitigate packet loss on the WAN, not to introduce loss by dropping VoIP traffic because another flow is trying to eat through the bandwidth.

The simplest way to prevent serialisation is to throw more bandwidth at a problem. The best way is to apply QoS, with maximum delay parameters set on multiple leaky-bucket classes of traffic. A packet spends too long in a queue it gets dropped, the VoIP / real-time / interactive queue gets emptied first.

Incidentally I can assure you that 20Mbit/s doesn't start to feel so nice when you have an unmanaged link and one person hammering at it with downloads. It is precisely to avoid such issues that I have 2 load balanced FTTC lines.

In reply to a post by Fastman2:

igntiionnet- there is no case for widescale FTTP (its just hard and complcated and expensve and indivually bespoke - it only really works in newbuild when tou can deployit as a greenfiels site) - i sure you are aware of the payback on commercial network -- and BDUk is even longer -- their is no case for FTTP

There seems to be a case in Milton Keynes for FTTP. Indeed despite many problems with the deployment it's still being pressed on with. There are certainly pockets around where it's apparently being viable. MK was apparently worthwhile as uptake is expected to be high.

I'm aware of the payback on the commercial network, hence my frustration at the delay over my own cabinet. It's probably going to be paid back in half the average time.

I guess the powers that be really got their sums wrong when the 25%, or 4.75 million premises, being FTTP in the early stages of the NGA deployment was mooted.

In reply to a post by Fastman2:

openreach still have to provide voice serivices as part of its operating licence so lots of FTTP and remove the opper sounds great in tehory - but openreach has to operate ithe real world and it has to be proven to be robust (so reality is another challenge all together)

I take it the tens of millions of VoIP lines running over HFC, or the few million FiOS telco lines running on battery backed up ONTs aren't considered examples of how robust a voice network could be?

If that is the argument you guys are in real trouble given those few streets that are Ebbsfleet have no copper. This doesn't work for me, if it did Openreach would've installed copper in Ebbsfleet next to the fibre just in case.

In reply to a post by Fastman2:

can you image the furore of your house buring down and not being able to place a emergency call could not be placed due to voice network not operating

Two problems there:

1) If my house were burning down I wouldn't be inside phoning the fire brigade, I'd be outside.
2) How many people now use corded phones? Most are on cordless which, just like FVA, require mains power.

In reply to a post by Fastman2:

-- (copper provides Voice) and wil continue to do so for a considerable period of time (i agree that FTTP is a preference in in new sites but then if you dont do any copper you have to deploy FVA Voice and there is only one provdier deploying that at present - that is also dependant on the develeoper just deploying copper as you know to your cost

There's only one provider of it because it's so ridiculously priced and FTTP availability is so poor.

If there were actually a concerted effort to drive FTTP doubtless more retail offerings would come into play. It's not a difficult thing for an operator to offer, it's largely the same as WLR just with a VLAN on a fibre link instead of a copper loop between premises and exchange.

Perhaps if Openreach actually made some effort to press the case for being able to retire the copper network something would happen. They seem able to, quite vocally and publicly, press the case for things that don't potentially threaten a high profit revenue stream such as the copper loop.

It costs about the same to rent a VLAN on a pre-existing fibre line for FVA as it does to rent a copper loop via MPF. Crazy. The addiction to copper within Openreach is made abundantly clear looking at the 'transitional' products. Penalising people for not wanting to keep a copper line active, while Verizon actually want the copper gone.

Openreach and BT Group could, if they wished, lobby Ofcom for variances to their license. Moving away from copper between premises and exchange would lower Openreach's maintenance bill considerably, though again it would also potentially lower the lucrative market in continuing to sell metallic loops that've been amortised time and time again.

Not sure how many houses have actually been built at Ebbsfleet, Google satellite still looks the same as when I visited the site four years ago.

Fibre Voice Access and battery more than fit the voice requirements, and if there was a fire in a home, you don't hang around to ring fire bridge you get out.