Print

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

[WWWombat]

Hmm second re-sync. Larger drop in upstream this time. Higher Latency has returned. Not so good...

It looks like DLM intervened upstream, and put an 8ms delay, INP=2.5 setting in place.

In turn, it looks like you had a lot of upstream ES's yesterday, with two prolonged bursts of CRC's at 6am and 7pm. Unfortunately, the CRC graph makes them look dwarfed by a couple of tall, one-off spikes downstream ... which aren't so important to DLM, as they only resulted in a couple of ES's.

You had similar prolonged bursts on Friday too. Heating?

[skyeci]

Could be. The weird thing is pre g.inp my upstream was always synced at 20mb even though the ds was a bit hit and miss. I seem to be in a worse position now than before g.inp...

[shadow4dog]

My BT line with a HH5a resynced today, on the same CAB. It's sync has increased from 61Mb to 65Mb, but it doesn't appear to have G.INP turned on (using the profile/sync rate calculation) It has no interleaving, appearing to be running in "fast" mode.

I'm gonna see if replacing it with an 8800 will make any difference... I assume the CAB will detect a modem capable of support G.INP and turn it on.

Tim

[blue166]

Just quickly, are the ECI modems working OK with G.INP on ECI cabs?

Cheers,

[skyeci]

Hi Tim

Which cabinet are you on? - I use an 8800nl which is detecting g.inp on my line..

Cheers

[WWWombat]

My BT line with a HH5a resynced today, on the same CAB. It's sync has increased from 61Mb to 65Mb, but it doesn't appear to have G.INP turned on (using the profile/sync rate calculation) It has no interleaving, appearing to be running in "fast" mode.

I assume your BT line isn't the one monitored by MDWS. Can I ask what we should call the monitored one, for future reference?

I'm gonna see if replacing it with an 8800 will make any difference... I assume the CAB will detect a modem capable of support G.INP and turn it on.

I can't be sure, but I thought the way it happened on the original Huawei rollout was that detection of valid modems was down to DLM - so wasn't a "live" decision. Therefore swapping between valid and invalid modems caused temporary mismatches of capabilities until DLM caught up. Perhaps that was amplified because, during the rollout, DLM took a while to catch up.

Perhaps this is a figment of my memory, or went away when the mk.II rollout happened. Anyone else remember?

[shadow4dog]

Hi Skynewb and WWWombat.

The cabinet is an ECI cabinet in Seaford, East Sussex.

I assume your BT line isn't the one monitored by MDWS. Can I ask what we should call the monitored one, for future reference?

Your assumption is spot on. It's my Plusnet line. They load balance between each other. I did have an HG612 attached to the BT line but it synced at a much lower rate - about 55Mb vs 60Mb for the HH5.

What I find interesting is the Plusnet gained 2Mb with G.INP enabled and the BT line gained 5Mb (was 60Mb now 65Mb) with no G.INP!

I think you're gonna be right... it's going to take a while for the DLM to notice it's not connected to the HH5 anymore.

Tim

[S.Stephenson]

Hi Skynewb and WWWombat.

The cabinet is an ECI cabinet in Seaford, East Sussex.

I assume your BT line isn't the one monitored by MDWS. Can I ask what we should call the monitored one, for future reference?

Your assumption is spot on. It's my Plusnet line. They load balance between each other. I did have an HG612 attached to the BT line but it synced at a much lower rate - about 55Mb vs 60Mb for the HH5.

What I find interesting is the Plusnet gained 2Mb with G.INP enabled and the BT line gained 5Mb (was 60Mb now 65Mb) with no G.INP!

I think you're gonna be right... it's going to take a while for the DLM to notice it's not connected to the HH5 anymore.

Tim

I've experienced the same thing with the exact same line, it gained 2mbps with the HG612 but with the HH5a it gained 5.

Interesting thing is even when I swap between the two G.INP still hasn't been turned off on the HG612.

[WWWombat]

Are you judging that G.INP isn't active (in some cases) based purely on looking at the IP profile ratio?

Personally, I wouldn't be so quick to judge on that alone. We already know that the outcome of ECI G.INP exhibits some visible differences, and I wouldn't be surprised if there were some invisible ones too.

I would say the technique has the potential to give us the answer, but like most things with the HH5A and the ECI modems, the details for the proof is lacking.

I'm trying to use the detailed stats to see if I can see more, so I'd welcome more examples...

[skyeci]

Hi,

Just wondered if someone could clarify that since my last re-sync according to the traffic light system everything is now back at 86400s and 0es so would there be any possibility of a re-sync upwards as my upstream has come down from 20mb to 17mb or is this the way dlm will leave it - just wondered if you are at the top end will it try to increase your speed etc.. My upstream does have delay,interleave and g.inp....

cheers

[WWWombat]

would there be any possibility of a re-sync upwards as my upstream has come down from 20mb to 17mb or is this the way dlm will leave it

All those errors that caused the bursts of CRCs and ES's (upstream) that triggered the DLM intervention have now become FEC's, leaving a zero ES rate. That means DLM has deployed a solution that successfully defeats the noise, but it doesn't mean the noise itself has gone.

I would imagine that DLM could choose to de-intervene if it only looks at the ES rate. However, my gut feeling is that it probably also checks the FEC rate, to see if that isn't too high; this would make sense to me, but has no official confirmation whatsoever.

Unfortunately, the MDWS graphs for FECs auto-scale themselves, and the sporadic, but large, spikes of downstream FECs make the behaviour of the upstream FECs almost invisible, so it is hard to tell.

My upstream does have delay,interleave and g.inp....

Your upstream now has had an old-style DLM intervention, setting an INP of 2.5 and a delay of 8ms; the modems have turned on FEC and interleaving. However, it doesn't look to have G.INP running upstream; I wouldn't expect it, either.

[ktz392837]

btw I updated the list to add MikeZ's stats and attached it below.

Any chance of an updated sheet? Very useful.  Any clues on why some users are loosing quite large amounts of speed? Tia

[tbailey2]

Unfortunately, the MDWS graphs for FECs auto-scale themselves, and the sporadic, but large, spikes of downstream FECs make the behaviour of the upstream FECs almost invisible, so it is hard to tell.

The graphs have always had a facility to drag/zoom any area so if you do this between any unwanted peaks....

[WWWombat]

The graphs have always had a facility to drag/zoom any area so if you do this between any unwanted peaks....

I was trying that recently, but it looked like you could only zoom so far, before it just chose to shift the focus forwards or backwards in time. Every change I made ended up with a large peak at the left, or one at the right; I could never focus on just the in-between portion.

[WWWombat]

I've been back and tried again. I was trying to replicate your image, and can only achieve the one attached below. I cannot zoom in a way that gets rid of both spikes.