DLM is broken. Explain this please.

http://i.imgur.com/BLa5ina.png

In the early hours of this morning after a week of a clean 71mbit sync. BOOM! It dumps me back on interleave of 1250, resyncs the line, puts INP on and starts doing the same old carp as before with RS_DS peaks..

BUT!

As soon as it turns this on it shows DS_RSCorr and DS_RSUnCorr as having spikes where it was clean before it kicked in.

Now the question becomes WHY is it kicking in? Why is the line quality WORSE when it turns on?

here is a 16 hour capture of the stats to see more detail when it kicked in. As you can see nothing major right before DLM decided to turn it's carp on.

http://i.imgur.com/BsQHwwl.png

Now on the previous graph you could say that it saw what happened on the 23rd (that huge spike in errors) and made a decision based on that but I have not been disconnected in the last week so WHY the hell is it turning on? The line was plenty stable as far as I can tell.

It's so frustrating!!!

# xdslcmd info --stats
xdslcmd: ADSL driver and PHY status
Status: Showtime
Retrain Reason: 2
Max: Upstream rate = 18475 Kbps, Downstream rate = 75512 Kbps
Path: 0, Upstream rate = 20000 Kbps, Downstream rate = 65719 Kbps

Link Power State: L0
Mode: VDSL2 Annex B
VDSL2 Profile: Profile 17a
TPS-TC: PTM Mode
Trellis: U:ON /D:ON
Line Status: No Defect
Training Status: Showtime
Down Up
SNR (dB): 5.9 4.9
Attn(dB): 0.0 0.0
Pwr(dBm): 13.3 6.9
VDSL2 framing
Path 0
B: 51 237
M: 1 1
T: 64 45
R: 12 16
S: 0.0252 0.3782
L: 20337 5373
D: 1287 1
I: 64 127
N: 64 254
Counters
Path 0
OHF: 42821694 507420
OHFErr: 203 287
RS: 2372262313 2427423
RSCorr: 1330802 2592
RSUnCorr: 5464 0

Path 0
HEC: 957 0
OCD: 31 0
LCD: 31 0
Total Cells: 162868949 0
Data Cells: 34093420 0
Drop Cells: 0
Bit Errors: 0 0

ES: 9189 2003
SES: 30 0
UAS: 38 38
AS: 69267

Path 0
INP: 3.00 0.00
PER: 1.61 4.25
delay: 8.00 0.00
OR: 119.16 60.17

Bitswap: 13128 4271

Total time = 1 days 22 hours 38 min 48 sec
FEC: 1330802 37586
CRC: 203 2795
ES: 9189 2003
SES: 30 0
UAS: 38 38
LOS: 5 0
LOF: 5 0
Latest 15 minutes time = 8 min 48 sec
FEC: 5070 19
CRC: 5 6
ES: 1 4
SES: 0 0
UAS: 0 0
LOS: 0 0
LOF: 0 0
Previous 15 minutes time = 15 min 0 sec
FEC: 18149 24
CRC: 0 9
ES: 0 7
SES: 0 0
UAS: 0 0
LOS: 0 0
LOF: 0 0
Latest 1 day time = 22 hours 38 min 48 sec
FEC: 1330802 2810
CRC: 203 322
ES: 134 249
SES: 5 0
UAS: 18 18
LOS: 5 0
LOF: 5 0
Previous 1 day time = 24 hours 0 sec
FEC: 0 2490
CRC: 0 416
ES: 988 293
SES: 7 0
UAS: 0 0
LOS: 0 0
LOF: 0 0
Since Link time = 19 hours 14 min 27 sec
FEC: 1330802 2592
CRC: 203 287
ES: 43 223
SES: 0 0
UAS: 0 0
LOS: 0 0
LOF: 0 0

Edited by Oddball (Sun 25-Aug-13 23:03:59)

In terms of explaining the graphs, I can't, especially as not being an expert on them I can't read any of the words.

But when interleaving is off you don't get any spikes for RSCorr and RSUnCorr because they are counters for the errors corrected or not by Reed-Solomon routines which require interleaving to be on in order to work .

When it's off, you just get retransmitted packets instead of any corrections.

Edited by RobertoS (Sun 25-Aug-13 23:00:27)

Hmm. So basically it will only show those spikes on those graphs because interleave is on and not if off (even though they are there?)? Sounds fishy. I just don't understand why it's decided to resync the line and put interleave back on. Going back those stats it looks the same for the last week apart from that big spike on the 23rd. But surely it would not slam on the brakes just for that one period? If it does then the algorithm really really sucks. Now if it caused the line to disconnected several times or that spike lasted for several hours throughout the day up until DLM did it's early hours 'let's see what is going on in the last 24 hours' thing then maybe I could understand why. But that spike lasted about 2-3 hours on the 23rd during the evening from what I can see. If the algorithm is basing that one period of bad quality on a period of more than 5 days of GOOD quality and slamming on the breaks then it is a rubbish algorithm!

Edited by Oddball (Sun 25-Aug-13 23:14:57)

In reply to a post by Oddball:

Hmm. So basically it will only show those spikes on those graphs because interleave is on and not if off (even though they are there?)?

That's right.

I just don't understand why it's decided to resync the line and put interleave back on.

As I said, I can't help with the graphs, but say you have 1000 packets with 400 error packets when interleaving is off.

That means you have to have those 400 packets retransmitted. So a total of 1400 packets instead of 1000. Bout of course, if the error rate stays the same, 160 of those retransmitted ones with again have errors. Then retransmit those 160 a second time, and 64 of those will have errors ....

But you don't know that is happening, except your throughput speed drops dramatically.

With interleaving on, and RS Correction running, you will get something like RSCorrected 395 of the 400, and RSUncorrected only 5. So 5 retransmissions and a fair chance they all get through without error. A total of 1005 packets transmitted instead of well over 1600, to get 1000 valid ones. Throughput speed is near normal.

Looks like a good reason to turn interleaving on to me .

Edited by RobertoS (Sun 25-Aug-13 23:29:55)

To be honest I don't care for it as my ping goes up and my download speed goes down. If I could have the damned thing turned off entirely I would. I have a feeling this is *NOT* to do with my line condition (or is only part of the equation), but to do with what it calculates based on what the other lines 'OVERALL' in the cabinet are doing (to prevent crosstalk).

I think it basically adjusts to the 'whole' as it see's fit to make sure everyone is getting a fair speed. That is purely my guess of course. It's probably more complicated than that.

I can't wait to see if vectoring gets rolled out and how it affects things.

I'm very selfish so I could not give a [censored] about anybody elses lines

I just want to stay synced at the same speed unless I can see something obvious going on with my line. On ADSL2+ I could see when noise caused my line to become unstable. On this system I can't seem to see anything that explains it.

On ADSL2+ I could see when noise caused my line to become unstable.

How? And what did you do about it when it did?

Edited by RobertoS (Mon 26-Aug-13 00:01:51)

Leading question. I can read between the lines

Basically monitored it using my Netgear router modem and some software that provided realtime stats (routerstats).

What did I do about it? Nothing I could do if I was forcing SNR to 3 instead of 6. Maybe raise the SNR if it was disconnecting me too frequently. But it always reconnected around the same speeds and I did not mind the occasional disconnect just as long as I got the max speeds I could out of it.

But in this instance I can see the speeds change and have no idea why and no control over it. I feel like I am back to the ADSL v1 days in that respect.

I'm OCD with this kind of stuff as you can tell.

Edited by Oddball (Mon 26-Aug-13 00:06:41)

burst of errors late on the night of the 23rd done it in for you.

whether your line is interleaved or not doesnt affect crosstalk, only the output power does.

Edited by Chrysalis (Mon 26-Aug-13 05:41:13)

Save up for fibre on demand now and say goodbye to this source of OCD pleasure

There is still an IP profile on FTTP Infinity.