In reply to a post by AndyHCZ:

I guess it came down to cost at the end of the day.

And getting it wrong [chuckle].

In reply to a post by AndyHCZ:

I wonder if Openreach are deploying 10G-PON for those ordering the fast speed variants?

That's my suspicion, too. BT did say, after multiplexing GPON with 10G-PON and NGPON2 (40Gb) that they could start offering symmetric services with that.

It may also be the case that they use GPON with a much lower split ratio.

In reply to a post by AndyHCZ:

You have evidence there is demand?

On the contrary, the average demand tells a completely different story...
https://postimg.org/image/sgsimrlo7/

That was 7:1 downstream:upstream.

Whose figures are those? An average of 190GB with 75% of data consumed by 25% of lines is probably rather skewed. It's a case of the mean not being the most relevant figure. The mode I think is what we need to see, plus the distribution graph of number of users v usage.

Chrysalis

All to do with scale and economics

The VLANs go all the way from the head end exchange to the BRAs in a core site. Thus you have Bandwidth in different physical paths with more and more VLANs from different sites in the same path the closer to the BRAs site you get.

So a 200Mb VLAN from Head end to Metro may run alongside other Head end on the same site. (You can also have other wavelengths in the same fibres from other sites carrying multiple VLANs).

Each Ge/10Ge will use a port on a edge router ( May be multiple edges at a site!) at the Metro, these get routed onto a Metro-Core path (10Ge/40Ge/100Ge) with lots of other VLANs.

Each VLAN can be rerouted at the Metro onto different paths by software.

This reduces the vast amount of potential oversupply that would occur if every Headend had a Ge/10Ge path all the way though the network to the BRAs ( including all the ports that would be only partially used at each Metro/Core/BRAs).

This whole design is why BT are so good (brilliant) when there is a fault at rerouting everything. It is also why capacity issues on VLANs are minimal and not usually seen by Customers. The scale also increases network resilience to equpment failure and enables service at a lower level to be continued rather than complete failure.

Due to the sheer number of VLANs etc most capacity issues are dealt with by robotic software and only when an additional Ge/10Ge needs to be provided are people involved to do the physical connection at the headend and the Metro. Even WDM can be set up remotely across the network to carry additional wavelengths when this is required.

AndyHCZ

VLAN capacity is managed by Robotic software within the Backhaul capacity at each site. As the base Physical capacity is 1Ge at a site new sites are not usually an issue for FTTC. If a new Haedend is likely to cover a lot of customers they may start with a 10Ge connection but this would be rare nowadays.

It is only when a VLAN is growing rapidly ( Faster than the lag to provide a new Ge/10Ge) does this get difficult (for FTTC). Simple forecasting tools can project total bandwidth used in a Ge forward to get a date when additional is required.

The delays on Ethernet services I believe are due to the work required to reach the customer site not the provision of capacity in the Backhaul network.

So Ethernet services for CPs including BT Wholesale do not have such delays.

In reply to a post by RobertoS:

Whose figures are those?

Sorry, I meant to say that...

It was one of the earlier snippets from Openreach's fibre index, and refers to the volumes on their NGA connections.

That one was summer 2015, and is no longer available on the OR website. Six months later (Winter 2015/16, and still available on the website), the total was 230GB, and the ratio remained at 7:1.

The current snippets show the total having reach 260GB, growing at around 25% per annum, but there isn't an update to the ratio.

In reply to a post by RobertoS:

An average of 190GB with 75% of data consumed by 25% of lines is probably rather skewed. It's a case of the mean not being the most relevant figure. The mode I think is what we need to see, plus the distribution graph of number of users v usage.

I agree that those would be more helpful, especially seeing quartile or decile distributions (perhaps debating whether median or mode is most appropriate to this dataset), but the data we're given data isn't entirely unhelpful.

First, we can take their 25/75 split, and turn it into 2 new averages:
- From 100 lines using 19,000 GB, calculate the usage pattern,
- Lowest 75 lines use 4,750GB; Highest 25 lines use 14,250GB.
- The 75% of lower-usage lines average 63GB
- The 25% of higher-usage lines average 570GB

Most reports that break the average down (eg one of the old Ofcom annual ones) show a hockey-stick pattern, so we can probably guess that the median is not far below the 63GB value. I doubt there's one "mode" value that stands out, but there's probably a lot of premises in the 40-63GB region.

That's quite a difference between the two sub-groups. My suspicion is that the difference is mostly explained by video:
- That the first sub-group, the 75% low-usage group, make use of Youtube and some iPlayer; the free video services.
- But the second sub-group is the set who have made the jump to subscription video services (Netflix, Amazon Prime, NowTV), and consequently use the subscription more.

If I'm right with that suspicion, I'd say there a pretty good chance that their downstream:upstream ratio is even higher than 7:1.

There was news earlier last year (about the time BTOR reported that Winter 2015/16 figure) that Netflix dominated UK subscriptions, with just under 25% of UK homes. Amazon Prime had about 6% and NowTV had 4%. Is that number is fairly significant.

There's obviously a question here: Is there a portion of the 25% doing something with even higher volume than watching video?

To help answer, we can make use of another piece of data to help here...

A while ago, a researcher for BT gave a presentation on usage volumes. He showed a graph of the usage seen on the busiest 20,000 lines in the country for one evening ... and the bulk of the least busy did indeed show video to be the main culprit.

However, the top 10% of those lines were primarily busy on other things, and consuming more bandwidth. So there is a niche group of users who are "busier on non-video"

The "top 10% of the busiest 20,000 lines" are perhaps 2,000 lines. At the time (mid 2014), BTOR had 3 million NGA subscribers, so the graph represented the top 0.6% of all NGA lines while the niche group of "odd ones" represents the top 0.06% of NGA lines.

My overall conclusion is that, barring a very small niche group, the faster we want our lines, the more we choose to consume our bandwidth through video, and the more our down:up ratio skews towards downstream.

you can come to the same results with leased lines at business parks tho.

The majority of traffic in datacentres is upstream, the majority for everything else is downstream.

Thanks. A real eye-opener for me! Particularly the Netflix figure.

Though I must admit I was surprised by an elderly lady last year who couldn't understand how she was getting extra charges from BT on Infinity 1, for exceeding her 20GB per month.

Shortly before her husband died suddenly they had bought a smart TV while still on 5Mbps ADSL2+. Continued to use a PVR. Then soon after he died that went kaput as well so she started using catchup TV.

She says the TV kept stuttering so she rang BT to ask why, and was sold Infinity 1 with 20GB. Which did cure it, but two or three months later she was using it most of the time and getting extra charges for going over the 20GB. She rang them again and got a free upgrade to 40GB. It wasn't long before she was hitting that as well.

She moved to a smaller house late last year and I think is now on unlimited so no problems.

In reply to a post by Chrysalis:

you can come to the same results with leased lines at business parks tho.

I'm not sure I can parse that to understand what you really mean to say.

Nevertheless, I was answering a thread that talked specifically about consumer behaviour. I don't have data that tells us much about business behaviour, except for the BSG study on small business needs (answer: it depends on size & nature of business, but majority of businesses have lesser needs than residential).

The exceptions shown up by the BSG report - such as larger businesses, software companies, hotels - are the small percentage that make good use of leased lines. I've got even less data about their usage.

In reply to a post by Chrysalis:

The majority of traffic in datacentres is upstream, the majority for everything else is downstream.

Though I guess you've reversed terminology there... what is "upstream" for a consumer is probably data being shoved out of a datacentre as "downstream".