Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[broadstairs]

dont know if you guys recall but I have always had loads of upstream FECs on my line and its never caused an issue with DLM.

as for things happening around 9pm mark and carrying on for a bit before trailing off - seems an odd time for any kind of automated system to be triggered. Could be a bug I suppose, but bugs are not well known for being regular as clockwork every 24h (unless human error during setup is in the mix). From working at the other end of the pipes, that time period screams out to me "the time when the most streaming, browsing and porn watching is being done". food for thought.

Well I can assure you that on my connection that is the quietest time of the day ...... 

Stuart

[plexy]

LOL Im sure you aren't doing pron at 9pm but dont forget all the other wires that your pair sits with on the way to the cab. Thats their busiest time, im pretty sure

[Ronski]

Well my works connection went up at 21:00 and stayed up around the 78,000 mark all night, still there now.

[broadstairs]

Well my works connection went up at 21:00 and stayed up around the 78,000 mark all night, still there now.

That happened once for me since I went to G.INP. I decided to change to the latest F/W on the 8924 and rebooted (obviously) and since then it has always dropped back just around 22:00. I do still see the occasional burst of upstream FECs at other times but only a quick burst and they are very sparse.

Stuart

[plexy]

Well my works connection went up at 21:00 and stayed up around the 78,000 mark all night, still there now.

That happened once for me since I went to G.INP. I decided to change to the latest F/W on the 8924 and rebooted (obviously) and since then it has always dropped back just around 22:00. I do still see the occasional burst of upstream FECs at other times but only a quick burst and they are very sparse.

Stuart

So could still be down to heavy use. Have we considered DSM i or ii ? would be good to know if the upstream power changes during this time period.

[WWWombat]

Well my works connection went up at 21:00 and stayed up around the 78,000 mark all night, still there now.

I was taking a look at the lines with upstream bursts, and noticed this effect (the ongoing FEC count) on Ronski's work line.

I wonder what kind of percentage of all the RS blocks these FEC's coming from? My line seems to run at just over 600,000 RS blocks per minute; however the counter (in "--stats") resets to zero every 6-7 minutes. I just jotted down the RS and RSCorr counters for bearer 0 once a minute, each time DSLstats refreshed the "--stats" tab.

10:28   3,622,705   12,142
10:29   4,239,307   12,142
10:30     579,126   12,142
10:31   1,213,901   12,142
10:32   1,869,843   12,142
10:33   2,483,472   12,142
10:34   3,118,248   12,142
10:35   3,770,519   12,142
10:36      92,841   12,146

Could you take a similar snapshot of those values for your work line, Ronski? Just a few minutes will do.

[roseway]

I'll just add that I see the same pattern. The upstream RS count seems to wrap around at a number which looks suspiciously like 2^32 / 1000. The downstream counters wrap around at 2^32.

[WWWombat]

Does ateece on DSLWS look like they have 3db set as target for downstream?

The Downstream dropped to 3db but the Upstream stayed at 6db so it kinda looks like it to me, went from 55mbps to 66mbps so it's looking good!

I went to have a look at these, but the stats now only start on April 1st - the original drop to 3dB is lost, so I can't see anything else at the time of the resync.

However, the current graphs show that the attainable speed is about 8.3Mbps below the actual (of 66Mbps) when the SNRM is 3dB, and attainable is about 10Mbps below the actual when the SNRM is 2.5dB. Those attainable values strongly suggest the target is still around 6dB, and the line just happens to be running at a low speed.

Of course, we can still observe the line's behaviour with an actual SNRM of 3dB, to see what errors it is now incurring. That might give us some idea of how lines with a target of 3dB might behave in the future. Unfortunately, because the old stats are lost, we can't compare the errors seen now with a "before" picture.

So what do I see?
- The SNRM graph currently runs fairly steady at 3dB downstream, with notable troughs to 2.5dB on occasion
- The FEC graph shows low downstream counts when running at 3dB, but notably higher levels when SNRM drops to 2.5dB
- The ES and CRC graphs show little of concern at any time, with no obvious correlation to the times running at 3dB or 2.5dB.
- The G.INP TX and Corr graphs seem to show retransmission is in regular use.
- The G.INP Uncorr graph shows a few irregular peaks. Little to worry about, I think.
- The G.INP LEFTRS and Min-EFTR graphs equally show little of concern at any time.

The difference in the FEC graphs is interesting, when SNRM is either 3dB or 2.5dB. That a change of just 0.5dB can require much more reliance on FEC is telling. It makes we wonder whether BT will really "just" introduce a new margin of 3dB, or whether they will introduce levels of 3, 4 and 5 dB.

I await more evidence with baited breath...

[plexy]

I'll just add that I see the same pattern. The upstream RS count seems to wrap around at a number which looks suspiciously like 2^32 / 1000. The downstream counters wrap around at 2^32.

var init fail? 

[Ronski]

Could you take a similar snapshot of those values for your work line, Ronski? Just a few minutes will do.

I've attached a zipped spreadsheet with a couple of days data, spreadsheet shows the field names - you can import into the spreadsheet if you want.

[Ronski]

My constant FEC's are still there on the works connection, but I did have a large peak in them around 21:00 last night.

There's no US FEC's on my home connection, but a few DS.

[broadstairs]

My constant FEC's are still there on the works connection, but I did have a large peak in them around 21:00 last night.

There's no US FEC's on my home connection, but a few DS.

I notice that your work u/s FEC counts increase during the 21:00 - 22:00 timeframe above what seems to be your base line. I also took a look at your home connection as well to compare it with mine. Our SNRMs are quite different, yours are around 6db for both where mine is a lot higher for d/s.  Also my sync speeds are quite a lot higher than yours. I'm really just clutching at straws to try to see if there might be a reason why your home connection does not see these FECs. I can almost set my watch by the start and end times of the FEC bursts on my line.

Stuart

[broadstairs]

I've been through all the G.INP users on MDWS with ECI cabs and found 11 others with similar issues to mine and a couple of similar ones to Ronski's work where there is a solid level all the time and an additional number during the 21:00 to 22:00 period, I also found 4 users who do not have the problem on an ECI cab. Now I'm not sure if MDWS stores anywhere the cab f/w but I know DSLStats does get that data. Since I can not see any obvious correlation between all these users I was wondering if the users who do not have a problem are on different cab f/w to those who do. If this info is not on MDWS anywhere I'll list the users I've found who DONT have the problem and see what F?W their cab runs (hopefully).

Stuart

[WWWombat]

Could you take a similar snapshot of those values for your work line, Ronski? Just a few minutes will do.

I've attached a zipped spreadsheet with a couple of days data, spreadsheet shows the field names - you can import into the spreadsheet if you want.

I saw the post yesterday with no file, and slightly later, when you added the comment to say you'd removed it. For some reason, I didn't see that you'd added the file back again.

I'll try to look later.

Both of the phenomena that @broadstairs sought in MDWS are very strange. The precision of 9pm - 10pm on a wide array of hardware, widespread across the country, is very strange.

[Ronski]

Added the wrong file, then the zipped spread sheet was too big