Print

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

[les-70]

   My HG612 consistently gives the lowest error rates and my line generates enough errors for errors to be the significant factor in deciding which modem to use.   The ECI may be better or worse but if I can't see the ES I don't like using a device.

[Ixel]

It appears DLM has raised my upload speed banding, from 15Mbps~ to 19Mbps~. It looks like the theory of keeping the ES below the good threshold in order to raise the speed banding may be correct, as DLM didn't lift it previously so I guess the ES was above the supposed rate of 288 daily on the 'speed' OR DLM profile (though I suspect there's some mathematical formula that takes FEC and other factors into account - FEC wasn't much yesterday either). My ES yesterday was 203.

This may also explain why I was stuck on 60/20 for several months when I experimented with capping a while back. Eventually I must've encountered a day, or a couple of days, where ES was below the good threshold (factoring in line history and FEC count - according to that training document) and so DLM raised the profile to 80/20. If I can keep ES below 288 in 24 hours then I imagine DLM will intervene again in the next day or two.

[Ixel]

I've gone back to the ECI /r this morning. Although I can't currently log any statistics from it (due to the way it's connected up, unless I can telnet from the DSL-AC68U via the WAN port to it), not to mention the statistics last I recall aren't entirely reliable anyway, I will leave it connected for a week or so and see if DLM increases the speed banding.

EDIT: I've put the HG612 back on again for 24-48 hours as the ferrite toroid has arrived. I've done 5 turns. Statistics available on MyDSLWebStats under 'Ixel', though so far I can't say it has made much difference to the intermittent CRC's, but it's early days I guess. QLN on DS3 looks smoother however.

[Chrysalis]

I am back down to circa 150 ES day now even on weekdays with no sudden bursts of CRC.

The people across the road started digging holes and I thought next was concrete for the car park, but today they finished up and gone.  No car park built.  They even removed a small shed type building did some work there and put it back after, odd.

However it has also been dry the last 2-3 days so this also could be related to the weather.

[les-70]

  I notice one person AnnM has exceeded 1440/day on two unusually bad days and, with no resync yet, is still on fast path.  May have been local thunderstorms  --- if so some say the DLM recognizes all lines are suffering and ignores it.

[Chrysalis]

or its possible 2880 is the actual limit, what isp is AnnM on?

[les-70]

  MDWS does not show the isp on the AnnM line.   When I see rather more than just these two  excursions over 1440 on MDWS i will believe in the 2880 values.  Apart from that exception, so far, no one occupies the 1440-2880 territory.

[Ixel]

As soon as I put the ECI /r back on, 24 hours later DLM intervened and changed my banding from 49/19 to 49/20. I'm hoping it will begin stepping up the downstream speed banding next.

[Chrysalis]

interesting is that ronski exceeded both 1440 and 2880 yesterday but is still on fast path.

I wonder if he will be DLM'd tomorrow and its not considered that data yet.

[les-70]

  It looks like events in one hour took him to 4171/day.  Unless that was discounted due to it appearing on all the lines on the DSLAM, as in a thunderstorm, you would certainly expect action.

[les-70]

  He has been DLM'd this morning. We can't learn much from that as he went well over the possible thresholds.

[kitz]

Over the past couple of days I seem to have a few isolated episodes of SHINE.
I dont know what these are about nor why they occurred.  From the times they occurred,  they weren't generated by anything I was doing.

What is interesting about these is the fact that the amount of CRCs generated yesterday should take me way over the MTBE red threshold if the DLM includes CRC errors.

We know that the DLM calculates MTBE normalised over the course of a day.  We dont know if that is midnight to midnight or any 24hr period ie ie current time minus 24hrs.

If the DLM monitors over the course of a day (say midnight to midnight) and uses CRCs, then in theory it should have taken action this morning.  If it monitors over any 24 hr period then it should also have kicked in this morning.

Ive had another burst a short while ago, not as bad as yesterday, but still sufficient to exceed the MTBE if its monitors over any 24 hr period.
Wondering how it will react now that it's seen two consecutive days whereby a 24hr period has exceeded the MTBE (if it looks at CRCs).

[Ixel]

I thought that the DLM should observe ES, not CRC errors. It apparently observes FEC, but in what way I haven't a clue. I imagine it either looks at the FEC seconds or the total FEC's during the 24 hour period?

[Chrysalis]

What is SHINE? forgive my dumbness

A day or two ago I had a burst of ES, thankfully I stayed below the ES threshold.

I believe kitz only ES is used not CRC, a short time ago I posted I had a 17k error burst and was not touched by DLM.

Kutz I just checked your graphs and your MTBE isnt affected, your ES looks normal for your history.
I do see the CRC spikes tho and they are probably a burst of CRC in a very short time period.  Did your SES increase?

A line that gets no errors for 12 hours, then a burst of 5000 errors all in one second, and then no errors for another 12 hours will have a excellent MTBE dispite the high CRC count, whilst a line that gets exactly 1 CRC error every 20 seconds whilst would have have a lower CRC count would also have a much inferior MTBE.

[Black Sheep]

SHINE = Single High Impulse
PEIN = Prolonged
REIN = Repetitive


SHINE = A pain in the arse.