Print

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

[wolvesmad]

Well I had a resync last night at 2:10am - checked the stats and I'm on retx low.

Rebooted the Zyxel and I'm now getting 75Mbps throughput 

[j0hn]

Excellent result.

[tiffy]

Well I had a resync last night at 2:10am - checked the stats and I'm on retx low.

Rebooted the Zyxel and I'm now getting 75Mbps throughput 

You have been lucky with your "speed capped" time scale required for DLM intervention and DS G.Inp re-tx profile change from high to low profile, I have just carried out a similar exercise on my rather long VDSL-2, 40/10 service line which is capable of just under 40 Mbps DS synch, capped at 30 Mbps and took 14 days for DLM to re-synch with DS re-tx low profile.

Purely from observation of the 3 VDSL-2 lines I have stat's monitored and archieved 24/7 for a long time, I've come to the conclusion that the main parameter DLM monitors with respect to G.Inp re-tx low/high profile application/removal is FEC error rate.
I'am also of the opinion that DLM parameters may have been re-defined a few months ago as all 3 lines in question previously ran quite consistantly with G.Inp re-tx low profile and subsequent DLM changes to re-tx high profile did not correspond to any recorded, noticeable increase in any error rates including FEC's.

Some very knowledgable and respected forum patrons have also sighted G.Inp "LEFTRS" & "rtx_tx" parameters as factors which could influence DLM re-tx profile control, capping DS synch rate will obviously have an effect on these error rates so it's difficult to positively conclude if these are contributory factors in the equasion.     

[wolvesmad]

Any idea what the threshold is for FEC's before DLM intervenes and switches the line to retx high?

[tiffy]

Any idea what the threshold is for FEC's before DLM intervenes and switches the line to retx high?

No, afraid not, as I said, it's only my current opinion that DS FEC rate is the main factor regarding DLM G.Inp re-tx profile selection, it's very likely that other factors contribute to the DLM equasion to some extent.

Can state that on my line, PN 40/10 service, without modem DS speed capping my  "FEC error/hr" daily rates as logged by DSLStats has consistantly run in the millions range, with the DS 30 Mbps modem clamp applied dropped to low hundreds where it took 14 days for DLM to revert to G.Inp re-tx low profile.
Have increased my modem DS clamp to 35 Mbps today and will monitor/log the effect on FEC error rate the obvious objective being to get as high a DS synch speed as possible while remaining on re-tx low profile to obtain the highest possible BRAS IP profile.

On the other lines I monitor, for comparison:

BT 80/20 service, Huawei DSLAM, relatively clean line, very low DS/US error rates, current DS synch, 64.8 Mbps, G.Inp re-tx high profile, "FEC error/hr" average daily rate 10,000, no modem capping applied.

PB 40/10 service, Huawei DSLAM, quite a noisy line, current DS synch 35 Mbps, G.Inp re-tx high profile, "FEC error/hr" average daily rate, 50,000, no modem capping applied but believe that the line may be DS banded @ 35 Mbps.

[tiffy]

Having gradually increased the DS modem clamp on my PN 40/10 VDSL-2 service over the last few weeks in 2 Mbps. steps at approx. 7 day intervals up to 40 Mbps. while monitoring/recording DS FEC error/hr daily rates via DSLStats and still assuming that FEC error rate is the main  parameter monitored by DLM with respect to G.Inp low/high re-tx profile application/removal can report:

Obviously and to expectation, DS FEC error rates decrease in proportion to the level of DS synch speed reduction produced by modem clamping.
As previously reported, found that quite a considerable reduction of DS FEC error rate produced by application of a low modem DS clamp was initially required before DLM action applied re-tx low profile, in my case this took 14 days, others have reported a much shorter time scale.

With re-tx low profile now active the DS FEC error rate of course increased with every 2 Mbps modem DS clamp increase, by approx. 2000 error/hr daily rate in my case but still maintained re-tx low profile on every modem re-synch after clamp changing and no DLM executive action recorded outwith planned re-synch's.
Noted that the FEC error rates were inclined to increase at week ends, attributed to greater local cabinet activity and ignored.

With the modem DS clamp now at 40 Mbps my DS FEC error rate is averaging approx. 8000 error/hr daily rate which is certainly much lower than previously when DS re-tx high profile was active with no DS modem clamp applied and DLM so far has retained re-tx low profile.
Considering my contracted service is 40/10 and my line would only ever be capable of marginally over 40 Mbps DS, can't really explain why the 40 Mbps DS modem clamp has the effect of considerably reducing my DS FEC error rate as it does ?

However, quite happy with the current situation and will continue to apply the modem DS 40 Mbps clamp in the hope that re-tx low profile will be retained by DLM, my current DS synch/BRAS IP ratio is 96.69%.

[ejs]

Probably the difference in FEC rate at 40Mb is due to the different rtx low/high, not due to the 40Mb rate cap set on the modem.

[tiffy]

Probably the difference in FEC rate at 40Mb is due to the different rtx low/high, not due to the 40Mb rate cap set on the modem.

Many thanks for the input, however I'am confused.

Accepting that FEC error rate is very likely one of the factors that DLM monitors with respect to low/high re-tx profile selection and high rates appear to trigger re-tx high profile application as DLM tries to improve the error rate, this would indicate (to me anyway) that a re-tx high profile will always tend to produce a lower FEC error rate than a re-tx low profile otherwise why would DLM apply a high profile ?

From observation of my line stats over a long time period in both DS re-tx low and high profile states don't think that the DS synch rate as such changes significantly due to DS re-tx profile state but the BRAS IP profile of course does, approx. 97% on re-tx low and 91% of DS synch rate on re-tx high profile, this will obviously effect the max data through put rate.

The currently applied 40 Mbps DS modem clamp should really be hypothetical on my line being 40/10 contracted but it does seem to keep the DS FEC error rate lower, considering the observations as recorded above and with the line being currently on DS re-tx low profile, the DS FEC error rate surely has a greater potential to be higher than if it was running on re-tx high profile ?

Not that I'am complaining, my line is running well with all error rates low, max possible DS synch and nearly 97% BRAS IP/DS synch ratio.

[ejs]

Have you tried removing the 40Mb cap on your modem, or setting it higher than 40Mb?

[j0hn]

Did you take note of all the framing parameters and check they were identical on both your measurements?
DLM can tweak a lot more than just Retx Low/high, also making INP, interleave depth etc adjustments.

Try again without the 40Mb cap and you should find it's the same amount of FEC.

Maxdatarate does nothing special and should behave exactly as a product cap or DLM banding.

I've ran with a cap above my sync speed and a cap at my product cap speed and never noticed any changes to the ES rate or FEC rate.
Only lowering the sync with a cap had any impact on error rates.

[tiffy]

Have you tried removing the 40Mb cap on your modem, or setting it higher than 40Mb?

Have tried this previously and found that the DS FEC error rate gradually increased and DS re-tx high profile was eventually re-applied.

@j0hn:

Did you take note of all the framing parameters and check they were identical on both your measurements?
DLM can tweak a lot more than just Retx Low/high, also making INP, interleave depth etc adjustments.

Yes, DS interleave depth did alternate 4, 16, 8 during the different capping trials, settled now at 8.
DS INP also decreased, now at 45.00.
Have always found that DS INP is 50 or less with re-tx low profile and 51 or greater on re-tx high profile, almost as positive a guide as bearer 0 INPRein 0/1 status.

You are absolutely correct of course, these are other parameters which need to be considered when evaluating DS FEC error rate with respect to DS synch rate modem clamping, I do have all the stats archieved and will get around to tabulating against specific clamp rates.
Will leave my modem clamp at 40 Mbps at present as the line is performing very well with very low error rates.

Many thanks to both for your interest and input.

Edit: Typo correction.

[tiffy]

Apologies to "wolvesmad", on reflection, I appear to have virtually taken over your post !
Still, hope this was deemed on topic and the resultant discussions have been informative with respect to your original question.

[tiffy]

Have always found that DS INP is 50 or less with re-tx low profile and 51 or greater on re-tx high profile, almost as positive a guide as bearer 0 INPRein 0/1 status.

Ironically, have just seen an exception to this theory on one of the lines I monitor, INP 49 and bearer 0 INPRein 1, re-tx high profile active.
Going back to the old MDWS days when large scale line monitoring was very easy have not seen this state before.

Another theory bites the dust !