Kitz ADSL Broadband Information
adsl spacer  
Support this site
Home Broadband ISPs Tech Routers Wiki Forum
 
     
   Compare ISP   Rate your ISP
   Glossary   Glossary
 
Please login or register.

Login with username, password and session length
Advanced search  

News:

Author Topic: BTs new DLM algorithms  (Read 5169 times)

sevenlayermuddle

  • Helpful
  • Addicted Kitizen
  • *
  • Posts: 5189
BTs new DLM algorithms
« on: June 22, 2010, 09:30:15 PM »

I notice BT have recently published another DLM patent, this time for a system that includes different stability options.  It may or may not describe BT's own DLM in current deployment (or for 21cn), but I'm betting that it does.

see: http://www.freepatentsonline.com/EP2169980A1.html

Click on 'download pdf' rather than trying to read the html version which doesn't render properly.

Like all patents, it looks like it's been written by an engineer and then encrypted into gobbledegook by their legal beagles, so reading it is hard going.
But it's probably worth the struggle as it's loaded with intriguing details.  For example, it explicitly defines, (table 2) various  thresholds for resync intervals, and time between errors, that are needed to get you penalised by a higher target, or rewarded by a reduced target.

I've yet to absorb it, and I may regret drawing any early conclusions, but to whet your appetites...

...It looks you need less than one error in 24 hours (86400 seconds) to eventually qualify for a reduction on the 'stable' profile, whereas one error in 2.4 hours is what's needed on the 'aggressive' profile.   I'm not yet sure what it means by 'errors'.

...Ten retrains in an hour always gets you naughty points, regardless of the profile.  This specifically includes power-off resets.

- 7LM

PS:  It's also largely about area-wide events, such as thunderstorms.  It seems that if a high percentage of lines in a given neighbourhood suddenly starting retraining frequently, as a thunderstorm is likely to affect many lines, then the retrains aren't counted for profiling.  So a thunderstorm should no longer get you sent to 15dB jail, no matter how many times you own line retrains.
« Last Edit: June 23, 2010, 12:11:08 AM by sevenlayermuddle »
Logged

roseway

  • Administrator
  • Senior Kitizen
  • *
  • Posts: 39663
  • Penguins CAN fly
    • DSLstats
Re: BTs new DLM algorithms
« Reply #1 on: June 22, 2010, 10:32:06 PM »

Thanks very much for that, 7LM. Definitely icepack-on-the-head stuff. :)
Logged
  Eric

waltergmw

  • Content Team
  • Kitizen
  • *
  • Posts: 2772
Re: BTs new DLM algorithms
« Reply #2 on: June 23, 2010, 07:56:15 AM »

Hi 7LM,

I don't suppose we'll ever find out (but I don't know why we shouldn't) what defines an area.
I can appreciate that a thunderstorm is hardly likely to restrain itself to individual area codes let alone individual exchanges in that area.
In our case in Ewhurst we have a spate of problems on one long D side aluminium cable, some of which are almost certainly triggered by human intervention, but most of the remainder is working within our usual mediocre levels.

Kind regards,
Walter
Logged

sevenlayermuddle

  • Helpful
  • Addicted Kitizen
  • *
  • Posts: 5189
Re: BTs new DLM algorithms
« Reply #3 on: June 23, 2010, 09:22:57 AM »

I don't suppose we'll ever find out (but I don't know why we shouldn't) what defines an area.

I think it's all lines sharing the 'aggregation point', which is suggested to be the DSLAM / MSAN.   I don't know how many lines are served by each DSLAM/MSAN.  I may be wrong, but I think it can be anything from 32 up to 10s of thousands, depending upon hardware.

Reading through it again, I'm interested to see that this DLM also attempts to ignore user-initiated retrains caused by unplugging phone wires or switching off router. It cross references another patent (EP 07255001) that I can't find anywhere, but paragraph [17] seems to suggest that if your line disconnects, and there's a delay before it reconnects, it assumes user-activity and ignores the event.  It occurs to me that this still wouldn't ignore manual retrains initiated by the router's GUI or CLI, as these would often reconnect immediately.

I'm getting the feeling that BT have addressed a number of failings in the old DLM, and tried to resolve them, which is probably good.  More pessimistically, I can't help wondering if they've tried to fix a bad design by making it more complicated, and that is nearly always a bad thing. :-\

- 7LM
« Last Edit: June 23, 2010, 09:27:07 AM by sevenlayermuddle »
Logged

jeffbb

  • Content Team
  • Kitizen
  • *
  • Posts: 2329
Re: BTs new DLM algorithms
« Reply #4 on: June 24, 2010, 06:25:09 PM »

Hi
Still burning midnight oil on this document.

Why use 5 words when a thousand will do ? :lol:

Regards Jeff
Logged
zen user

sevenlayermuddle

  • Helpful
  • Addicted Kitizen
  • *
  • Posts: 5189
Re: BTs new DLM algorithms
« Reply #5 on: June 24, 2010, 07:59:23 PM »

Hi
Still burning midnight oil on this document.

Why use 5 words when a thousand will do ? :lol:

Regards Jeff


Or even just a plurality of words where fewer would do.
 :D
Logged

BritBrat

  • Kitizen
  • ****
  • Posts: 1356
Re: BTs new DLM algorithms
« Reply #6 on: June 24, 2010, 10:23:39 PM »

I guess they must think the more words they use the more chance they have covered every angle.
Logged

jeffbb

  • Content Team
  • Kitizen
  • *
  • Posts: 2329
Re: BTs new DLM algorithms
« Reply #7 on: June 25, 2010, 03:38:06 PM »

Hi
Well at least I think I have found why It takes so long to improve (lower )target SNR .
Even-though the info is from the "new ?" DLM document it seems to explain the way it works now (or at least very similar)

This is a very much simplified view  .

The target SNR is based on the line quality monitoring  

It seems that there are 3 counters involved .Recording the state of the connection  Good delay  , Bad delay   and Delay Doubler
The good delay counter .This increments daily  up to a maximum of 13 .Starting at Zero when a target has been set . This gives a total of 14 days .If the good counter reaches 13 then the Target SNR may be lowered (depending on many conditions )

The Bad delay counter is set to 3 . Then if necessary the Target is raised(depending on many conditions )

The delay doubler initially set to 0. This increments to a maximum of 5 ,when A line goes  to a more aggressive state (higher synch , lowered Target SNR)

So if the target is lowered once the the doubler is at one , now you have a bad event causing the target to be raised . Now if the line is good , the good counter increments daily (hope) at the end of 14 days if all is well then the target SNR is lowered (if all conditions right). The Delay Doubler increments by  1 . Each increments doubles the delay before the target can be lowered again . So now the Good counter will have to run for 28 days  . The Good counter will have to cycle twice . And it goes on

So the affect of the Delay doubler is to double the delay time for each occasion of a reset to lower margin (yo yo action).

Delay doubler  set to    delay in days
                           0            14
                           1            28
                           2            56
                           3           112
                           4           224
                           5           448

before the target SNR target will be reset .

Regards Jeff

for a much more detailed explanation see section 0076  of the DLM document

Regards Jeff

edit  error in text




      


  

« Last Edit: June 25, 2010, 07:27:17 PM by jeffbb »
Logged
zen user

roseway

  • Administrator
  • Senior Kitizen
  • *
  • Posts: 39663
  • Penguins CAN fly
    • DSLstats
Re: BTs new DLM algorithms
« Reply #8 on: June 25, 2010, 03:55:21 PM »

That sounds entirely feasible, Jeff, thanks for putting it into comprehensible words. :)
Logged
  Eric

sevenlayermuddle

  • Helpful
  • Addicted Kitizen
  • *
  • Posts: 5189
Re: BTs new DLM algorithms
« Reply #9 on: August 25, 2010, 12:29:51 PM »

For those with remaining interest DLM, I stumbled upon this document. 

Search the document for 'DLM' and you'll find an embedded zip, from a BT customer workshop, containing detail that seems to spill the beans on some of the nitty-gritty details.  All in 'microsoft office' documents I'm afraid.

http://www.btwholesale.com/pages/downloads/21_Century_Network_Community/Monthly_summary_Industry_Engagement_May10.doc

You'll find some interesting plans for WBC DLM, and also some suggested inadequacies of ADSL1 DLM.

- 7LM

Logged

roseway

  • Administrator
  • Senior Kitizen
  • *
  • Posts: 39663
  • Penguins CAN fly
    • DSLstats
Re: BTs new DLM algorithms
« Reply #10 on: August 25, 2010, 12:57:58 PM »

That's going to be very interesting reading. Many thanks 7LM.
Logged
  Eric

DrTeeth

  • Member
  • **
  • Posts: 70
Re: BTs new DLM algorithms
« Reply #11 on: September 05, 2010, 01:22:36 PM »

I guess they must think the more words they use the more chance they have covered every angle.
They have to in patents. The more specifically defined a patent is, the easier it is to fond a loophole/work around. The cock-up that was the Epilady patent is legendary in the industry.

DrT
Logged
Stress - the condition brought about by having to resist the temptation to beat the living daylights out of someone who richly deserves it.
 

anything