Kitz Forum
Announcements => News Articles => Topic started by: Bowdon on October 25, 2017, 06:03:47 PM
-
https://www.ispreview.co.uk/index.php/2017/10/openreach-pause-plans-test-xdb-boost-uk-eci-fttc-broadband-lines.html (https://www.ispreview.co.uk/index.php/2017/10/openreach-pause-plans-test-xdb-boost-uk-eci-fttc-broadband-lines.html)
Last month we revealed that Openreach were preparing to test a ‘Proof of Concept’ upgrade that could boost the performance of ECI based FTTC (VDSL2) “fibre broadband” lines by adopting a default target downstream noise margin of 3dB (here). However the plan for this test has now been paused.
The XdB upgrade itself is a tweak to the SNR (Signal to Noise Ratio) of a DSL based broadband line, which reflects the balance (measured in decibels) between the useful information coming down a line (good signal) and unwanted interference (bad signal / noise). Previously Openreach only adopted a default target downstream noise margin of 6dB, but by dropping this to 3dB you could deliver a modest speed boost to stable lines (especially very short ones).
This is why I don't get excited these days by the stories of sending 1Gb down copper lines. Because when it comes to reality it doesnt work reliably.
I don't see point of having all these experiments. BT/OR need to focus on whats practical and how their products are going to move forward. I was expecting G.fast to be out commerially by now, yet apart from a page on TT's website page, nothing as been heard.
... and still nothing that Infinity 2 people are still paying at least £10 more for the same service as Infinity 1 people.. :'(
-
Glad they've dumped it .................. I've said it before, I'll say it again ............. for a very large percentage of lines, 3dB profiles cause problems.
-
No Surprises there, I'm getting mildly excited about getting proper fibre to my home and being able to dump fttc and ECI :P
-
Not surprising and probably a good thing to be honest. Hopefully G.INP is still on track to be rolled out nationwide in the near future (next year?).
-
No Surprises there, I'm getting mildly excited about getting proper fibre to my home and being able to dump fttc and ECI :P
Me too :P
Stuart
-
I think BT/OR need to accept that in the real world we're at the first edge of what copper lines can do.
I'll be watching how this G.fast roll out happens as I think G.fast will be the turning point in thinking.
-
No Surprises there, I'm getting mildly excited about getting proper fibre to my home and being able to dump fttc and ECI :P
Me too :P
Stuart
Believe me lads ....................... plenty of other folk are glad, excited you're off to VM. :P ;) :)
-
Glad they've dumped it .................. I've said it before, I'll say it again ............. for a very large percentage of lines, 3dB profiles cause problems.
Do we know what this large percentage is?
I'm quite surprised because 3dB works perfectly fine with ADSL2+ and on other SP DLM's without the need for G.INP It should only cause problems if daily SNR swings are more than a couple of dB.
Even in such cases, because of the way DLM works, it should soon learn not to attempt it again. There's a reason why DLM takes longer to relent at each subsequent step and that's to stop lines flapping. It shouldn't even try it anyhow unless the line is ILQ green, therefore problematic lines shouldnt get it anyhow.
There's several of us on here who have run lines at 3dB for weeks without g.inp and without any problems after power outages at the DSLAM. I've done so myself until I had to do a manual reboot.
-
Do we know what this large percentage is?
Fair do's ..... I can't quote numbers ..... allow me to quantify my wording.
For the most part, when I pick up a 'Flappy DSL circuit fault' and notice it is on a 'sub 6dB profile' ............. I immediately look at WHOOSH RRT. For the most part, the 'Flappy' started to occur the day it went 'sub 6dB'.
I'm not disputing certain lines will benefit from this profile, especially lines like yourselves, kitz, whereby you could throw a onion bhaji from your doorstep and easily hit your DSLAM ..... but for Mr/Mrs average who far prefer stability over squeezing every last ounce out of the DSLAM .... sub 6dB is not for them.
Point in case, my very own circuit ... which literally is spun from gold. Not had one issue since I reset the circuit and 'un-checked' the 'Sub 6dB' profile ... ergo stopping DLM attempting it again.
-
un-checked' the 'Sub 6dB' profile
This is one of the things I dislike about the NGA DLM in that they don't allow some control to the EU/ISP.
Wouldn't it be good if just like the 20CN/21CN system that this & Interleaving could be controlled at the ISP level. I'm sure it would also save you guys some time too especially if it came to resets - as long as ISPs were warned not to abuse it, then it would make things easier all round.
-
Completely agree ^^^^ :) :)
-
BS works in the field so i accept what he says about problems also that it seems logical it would be high risk without g.inp.
my real concern with this news is if it means the g.inp rollout has hit another barrier
-
logical it would be high risk without g.inp.
Perhaps I am being thick, but I really can't understand why it should be so problematic for NGA.
Its not a problem for other DLM systems.
NGA DLM needs a serious overhaul and rethink because as it stands its just botched together and too many parts of it don't work as it should. This is just further differentiating the 2 systems and making the divide between ECI & Huawei even greater. :(
The theory is pretty simple:-
You keep 6dB as default.
If and only if the line is ILQ green for a period of time, then you drop to 5dB.
If the line goes ILQ red then you immediately back out and return to 6dB
If the line goes ILQ amber then you take no action.
If and only if the line is ILQ green for a couple of days then you try 4dB
and so on.
In the meantime what was the ILQ 'doubler' method, but is now something like 2d,10d,24d, 8wk ensures that the line doesnt keep flapping.
-
I found it very interesting that BS revealed there is a off switch for xDB DLM. So there is at least 2 known mechanisms that allow override but are not made available outside of openreach.
To backup what black sheep has said I have friends who have told me when on 4db (never mind 3db) their line loses sync daily even tho ES and SNRM look stable. There is also the instances of lines flapping between different SNRM targets which should be fixable by adjusting DLM parameters so its harder for a line to move down to a lower target.
I have already accepted ECI service will never be on par with hauwei, you will just keep feeling let down if you build your hope up too much. End of the day it could have been much worse, we could still be stuck on ADSL, if I had a choice between ADSL, VM cable or ECI FTTC, I would choose the latter.
-
I found it very interesting that BS revealed there is a off switch for xDB DLM. So there is at least 2 known mechanisms that allow override but are not made available outside of openreach.
Yep me too.
To backup what black sheep has said I have friends who have told me when on 4db (never mind 3db) their line loses sync daily even tho ES and SNRM look stable.
So something isnt working as it should. What is triggering a resync if the SNRM is stable and without Err/Secs?
There is also the instances of lines flapping between different SNRM targets which should be fixable by adjusting DLM parameters so its harder for a line to move down to a lower target.
Again something isn't working as it should. The doubler system is supposed to do just that. (see my above post which I was editing at same time you were replying)
The doubler system had been extended so that times before allowing a retry gets longer and longer. Ironically we see some reports that it takes too long.
As I said above the whole NGA system needs a rethink.
-
I got no idea why their lines drop kitz, its only 2 friends, not loads to be fair, but it seems vdsl is not as solid as it should be.
-
I don't know, but if anything it seems like what was the doubler system seems borked. I think this is one of the things they tweaked after the ASSIA case, because its only since then we started seeing complaints about banding not properly being removed. The INP part appears to be working as intended.
When they first brought it out we seldom saw complaints about it. It was only after ASSIA... then later on with g.inp that things started to go out of balance.
We've also got the rather strange situation whereby Openreach will attempt to remotely remove a cap, but it never gets done right. Its only the Engineers resets that seem to work as they should. :'(
It's like they started with one system and have made too many amendments and something has got broken along the way.
The original NGA system was built based on banding and INP without Target SNRm.
Since then something has been changed with banding, the ILQ doubler has been extended. G.INP introduced. G.INP mk2 introduced.. and now its gone to using various target SNRMs.
-
I am convinced that there are more issues with DLM and I also suspect there are issues with it on ECI cabinets. Having managed to get my line running without a cap I still see issues with DLM refusing to remove interleaving despite (according to MDWS) my line being green. At its worst my line was left interleaved at 6db SNRM with very low ES and green status for 4 months with zero action by DLM.
Frankly I believe that DLM is not fit for purpose and I also believe it does have significant issues on ECI cabinets. If it cannot work properly with 6db SNRM how the heck can it cope with reducing the SNRM at all.
Stuart
-
On the theory that 3dB should work, and comparison with ADSL2+, where it is successful...
VDSL2 has two problems that DLM seems to not be designed for:
a) SNRM running below target because of DSLAM resets or power cycles
b) SNRM running below target because of crosstalk/takeup increases. Or even one neighbour who turns their modem off daily.
As we all know, running at a lower SNRM invites higher error counts, which can trigger DLM. Running lower SNRM targets makes this more likely than before. Higher takeup makes it more likely too. XdB and high takeup (without vectoring) have fundamentally changed the playing field since the days of 2013. We can no longer trust that a line runs within , say, 1dB of where we set the target. Not over a period longer than a few hours.
DLM patently shouldn't interfere when the running SNRM has badly diverged from the target. XdB requires DLM to be more lenient; probably only triggering a resync as the first response, and then only choosing a further step based on, say, the stats for 12 hours after that resync. Any other statistics are becoming less trustworthy ... such that GIGO rules.
-
On the XdB off-switch...
I'm still convinced that DLM got redesigned (after G.INP fiasco #1) into 2 layers:
- one operational layer, working as before, each night
- one monitor layer, that set the boundaries that the first layer worked in.
The supervisor would identify external conditions (such as modem type) to determine whether G.INP was a suitable quality-improvement tool, on this line, for day-to-day use.
I then think a lot of new features come with an inhibitory off-switch ... including G.INP and XdB. These switches force the monitor to discount that feature.
When we see a rollout take 3 months+, turning on N lines per day, I suspect that the underlying DLM process changed on day 1. The slow release is made to happen by turning the off-switch on for a few lines at a time.
In this model, each process has its own history for a line; the operational process changes daily, but the monitor is slower to respond. What we used to see as a reset of DLM can now happen in two ways, depending which histories get cleared.
With this model, I think banding changed ... such that it became part of the monitor layer, and became hard to dislodge. And immune to some DLM resets.
So... with such a model in mind, it is easy to think of multiple inhibitory off-switches existing, some controlled manually (as BS did), and some automatically (such as during a rollout, or by modem detection).
-
On the doubler...
Yes, that mechanism seemed to change ... and in some ways improved. I agree that the Assia case brought many alterations to the party, in a similar timescale to the consequences of the ECI fiascos. And it will have impacted the way line profiles were selected based on line history.
The problem for us is that, while identifying why DLM intervenes is hard, figuring why DLM de-intervenes is hard^2 or worse. Until we can insert a line simulator into a circuit, anyway !
-
From a personal perspective Xdb has achieved very little for me.
When I originally got VDSL it was pre G.INP and my line synced at about 35Mbps with a throughput of about 33Mbps.
G.INP then came along and my sync increased to the low 40's with an associated increase in throughput.
Xdb has now put me on a 4dB profile and my sync has increased to 46Mbps.
However, I have seen no increase in throughput as I assume I am now on a retransmission high profile which has pretty much killed any benefit from the increase in sync.
ES have gone from about 100 per day pre XdB to around 250-300 per day now.
:(
-
The 3dB margin significantly improved things on my backup BT line. The primary line remains banded and will be ceased in about a month.
I can't say I'm sorry to replace my primary xDSL line with something else either. It's worth remembering that Openreach started deploying this stuff in 2009.
-
tubaman that sucks if true, DLM should in my opinion if it needs to raise interleaving/retransmission to be stable at a noise margin then it should revert the noise margin and drop interleaving/retransmission back to where it was, so basically if you are correct you now have a higher delay penalty for nothing.
-
Retransmission high has a higher maximum retransmission delay, but while nothing is being retransmitted, the latency is unaffected.
-
its still a higher penalty than low retransmission.
ejs am I right that you are of the opinion that max sync speed wins over everything else? as you seem to be in favour of an aggressive xDB DLM.
-
What do you think the difference in typically latency between retransmission low and retransmission high profiles is?
ejs am I right that you are of the opinion that max sync speed wins over everything else? as you seem to be in favour of an aggressive xDB DLM.
No, and I'm not sure where you got this idea from. I'm pretty sure that whatever the DLM does, people will be complaining about whatever it did. The general consensus will be that the DLM is rubbish, and therefore any details about why it's rubbish, whether they are correct, incorrect or irrelevant, do not matter.
-
ok thanks for clarifying i asked as you responded to two recent posts i made on it.
i think its ok for people to moan about it as after all its a broadband forum :)
i dont think dlm is completely rubbish but also i dont think its perfect either.
the xDB stuff puts dlm in a weird place as its supposed to be stabilizing poor lines but this applies performance settings instead
-
Retransmission high has a higher maximum retransmission delay, but while nothing is being retransmitted, the latency is unaffected.
The extension of this is that, on a line running Retransmission High, there is an expectation for it to be retransmitting more blocks, and perhaps re-re-transmitting a higher proportion too with a higher count of re-re-transmissions.
Those will reduce the effective throughput of the line (the EFTR). Perhaps this is why the IP Profile gets set to a lower percentage - just in case retransmission is severe.