On FTTP, Congestion can happen in 2 places: on the PON, shared with 32 other premises, and on the backhaul that your connection is part of.

If the problem is in the backhaul:
The gradual ramp-up of speeds when you first connected are an indicator of congestion in the backhaul pipe, with its capacity being increased incrementally as software systems identified it was overloaded.

Perhaps it is overloaded beyond the point that software can improve anything. Perhaps it needs human intervention to rebalance subscribers... Have you talked to anyone at the ISP (BT Retail) yet?

Or perhaps the systems don't consider you in need of rebalance because your data doesn't officially exist yet.

If the problem is in the PON:
2.5Gbps shared amongst 32 premises should nicely invoke Openreach's congestion mechanism, which means you should never get below 75ish Mbps each. And that would require all 32 to both subscribe to infinity 3 or 4, and all 32 to try to max it out simultaneously.

I don't believe that...

Thank you for your insight.

In the last few minutes I've gained access to the building's Facebook group and it appears internet is completely out on a few floors, and someone reports being on the 50th day without a connection. So I guess I should count myself lucky - and that there is a much bigger issue here. And historically there were far more widespread reports of outages. And images of many Openreach vans outside. And petitions for Hyperoptic to be allowed to install.

I will contact BT but at the moment I am still waiting for them to contact me about the engineer no-show.

If there were large outages is it conceivable that there is reduced capacity and more than 32 people sharing the 2.5 Gbps while repairs happen?

Well here is something... I put my number into the service status checker to see if there were any reported issues and it appears there is an open fault report in relation to the connection which was opened on Friday - the day the engineer was meant to show up but no-showed. I haven't contacted them about this at all nor have I received any contact about that fault. And it doesn't show up on the My BT page.

Maybe they're onto it already!

Screenshot: https://cl.ly/37293Y0X030b

In reply to a post by WWWombat:

On FTTP, Congestion can happen in 2 places: on the PON, shared with 32 other premises, and on the backhaul that your connection is part of.

If the problem is in the backhaul:
The gradual ramp-up of speeds when you first connected are an indicator of congestion in the backhaul pipe, with its capacity being increased incrementally as software systems identified it was overloaded.

Perhaps it is overloaded beyond the point that software can improve anything. Perhaps it needs human intervention to rebalance subscribers... Have you talked to anyone at the ISP (BT Retail) yet?

Or perhaps the systems don't consider you in need of rebalance because your data doesn't officially exist yet.

If the problem is in the PON:
2.5Gbps shared amongst 32 premises should nicely invoke Openreach's congestion mechanism, which means you should never get below 75ish Mbps each. And that would require all 32 to both subscribe to infinity 3 or 4, and all 32 to try to max it out simultaneously.

I don't believe that...

I had a speed drop when I was on Infinity 4 a while back when my speed dropped down to about 45Mbps, so I phoned BT but was told its not classed as a fault until it dropped below 40Mbps, so I gave a huge moan and was moved (their words) and it was fine.

But yeah, for the connection to drop down as low as the OP it wouldn't be the 1st place you said, I also worked this out to the lowest you would get to the exchange would be 78.125Mbps and like you said all 32 connections would also have to have Infinity 4 and hammering their connection, which is very unlikely indeed.

So what bandwidth is being used for lines on Ultrafast 2 due to we have a guaranteed minimum speed of 100Mbps, so they must up the hardware speed because they cannot reduce the 32 way split.

I know my connection suffers with speed drops in the evenings and over the weekends show here where you can see that the single threaded has issues getting the full speed, where as the multi threaded has no issues.



But that is fine if I wanted to download a large file I could just split them up into several files and then download them all at the same time.

TBH I think the OP's issues are between the Exchange and BT's Network at fault here.

Paul

Edited by PaulKirby (Sun 13-May-18 02:11:44)

In reply to a post by Brookzy:

Thank you for your insight.

In the last few minutes I've gained access to the building's Facebook group and it appears internet is completely out on a few floors, and someone reports being on the 50th day without a connection. So I guess I should count myself lucky - and that there is a much bigger issue here. And historically there were far more widespread reports of outages. And images of many Openreach vans outside. And petitions for Hyperoptic to be allowed to install.

I will contact BT but at the moment I am still waiting for them to contact me about the engineer no-show.

If there were large outages is it conceivable that there is reduced capacity and more than 32 people sharing the 2.5 Gbps while repairs happen?

I would say it might be an issue at your exchange going to BT's Network due to you should never drop that low going to your exchange.

TBH if it was me I would phone up BT tomorrow to report this issue ,sure they might be aware of it, but they might not, and at the same time you could "enquire" about the engineer visit.

As for more than 32 people sharing the 2.5Gbps to the exchange, that's cannot be the case, the Splitter node splits up 1 fibre into 32 smaller fibres and a Splitter Node can take in 4 fibre to create 128 fibre connections.

Paul

In reply to a post by Brookzy:

Well here is something... I put my number into the service status checker to see if there were any reported issues and it appears there is an open fault report in relation to the connection which was opened on Friday - the day the engineer was meant to show up but no-showed. I haven't contacted them about this at all nor have I received any contact about that fault. And it doesn't show up on the My BT page.

Maybe they're onto it already!

Screenshot: https://cl.ly/37293Y0X030b

Ah, that makes sense and why your speed was dropping that low.

TBH I thought it might of been the hardware that provides fibre to your area that was at fault, something like where it chops up the block of 32 connections into segments, if that went doolaly it might of caused issues on your connections.

But it seems going by that image that its the connection between your exchange and BT's Network.

Yeah, maybe the enginner went to the exchange and saw the issue and reported it, or maybe the system detected an issue and create a fault report, I don't think that will be linked to your account or maybe it gets added once its fixed, not too sure.

Also you "might" want to remove the fault and ONT numbers from that image.

Paul

Leaving the numbers in was a calculated risk knowing that some guys around here have access to certain systems and might shed more light on the fault... Have removed it now.

Sure enough, now it's 2AM, I'm back to over 200 Mbps. It's got to be congestion.

Even if there were 128 people sharing the 2.4Gb it really wouldn't be a problem. The issue is going to be with the BT Wholesale network out from the exchange or the connection between Openreach and BT Wholesale.

I wouldn't be to surprised if it were the latter.

Brookzy

It sounds like it is not the GPON with the congestion but the Virtual path you are on from the exchange to the Core ( VLAN) these can be various different sizes used to be from 100mb ( for ADSL2+) to 850Mb ( max within 1 Gb physical) but are now likely to go bigger to fit 10GB physicals.

When they set up all the unused connections these are likely to be allocated to a small VLAN, as they won't be carrying anything other than test packets. When a customer order is completed they would be moved to that ISPs VLAN which will be at the size they pay for. ( Or a cable link to their network). This may be only 500Mb and will only be upgraded when it reaches a set % of capacity ( say 85% for 2 hours over 5 days) If several Infinity 4 customers are provided and all start to stream at 330Mb this will be congested until it is regraded ( Usually a software task) or the customers are moved to a different VLAN, as Paul was.

If your connection has only just gone live you may be on a 'small' VLAN, if it hasn't been completed in BT's systems you may still be on the original VLAN that may only be 300Mb.

Hope this explains your present situation. You may have to wait until the job is completed ( by the missing engineer!) or a couple of days before the systems automatically increase the VLAN size.

Only if the VLAN ( or VLANs) are at the physical pipe size will any human have to interfere to physically increase the pipe size, 1Gb to 2*1Gb or up to 10Gb this will take longer.

From your initial SpeedOf.me plots, there are two major takeaways:

1) Saw-toothing. This is the main 'fault'.

2) The 'trend' curve, whereby it accelerates from 0 to 213mbps before levelling off at around 160mbps.

The good news is I don't think it has anything to do with the physical connection, ie the fibre itself. The reason is you simply wouldn't even get anything close to the speeds you're getting, and the cutouts wouldn't be a recognisable pattern like the sawtooth it is and much more random.

Given the trend line is recognisable as well, this shows to me that's the GPON itself balancing load (this looks very normal if left to normal hardware to level load). If you run a SpeedOf.me in the dead of night, this curve should be higher. It could also be another (much more light-touch) traffic shaper and indeed is quite typical of a Fair Queue (FQ) plot, and indeed if two shapers were working on the same bandwidth you'd get weirdness like this as well, but again, I doubt it's shaper-related.

The bad news is there's clearly something misconfigured. While it could be shaping, I actually think this is an I/O issue, like a network driver or a diagnostic profiler such as rudimentary packet dumper that blocks network traffic while writing the dump to disk. This is what I believe is the most likely explanation, as every engineer has forgotten something like switching off a profiler before, and given this is BT we're talking about, the numbskullery is much higher than average.

Whatever it is, I agree with the others and it's pretty clear to me that this is an issue with the backhaul (between the Exchange and the Core) not in the last mile.