VDSL - weird packet loss issue

For the last 6 months or so, I've had a rock solid DSL connection, but the IPv4 and IPv6 connections have periodically dropped out

Zen have been unable to find anything wrong (I believe it's inside BTOR)

I was reminded that I'd setup BQM graphs and was somewhat surprised to see very high packet loss reports - and that they're following a sawtooth pattern repeating at about 6 hours. This pattern is almost unvarying for at least 6 months (That's as far back as I looked, could be more)

My Broadband Ping

5% packet loss is enough to cause issues, 20% is _extremely_ serious levels of packet loss and at 30% it's no real surprise that the PPP connection is failing occasionally

The link itself has been very lightly occupied for most of this time and the sawtooth doesn't reset when the router's rebooted (or when DSL resets or IP connectivity resets) - I'm pretty sure this is a BTOR issue but it'd be interesting to see if anyone else is suffering similar issues - particularly if you're in Surrey

==

Your using the frizbox, it has weird and wonderful ways of responding to ICMP

Ha! I too am seeing this. Zen FTTC, 54 from our Draytek router, rock solid for years and about 2 months ago, similar traces to you. I've tried a difernet Drayetk & also the Zen supplied Frits (horrid thing!) and all are prettty rubbish.

Zen have been v helpful and even OR who turned up were useful too. But despite best endeavours, no fix and the support case is still open.

Ongoing....

Yes, a 7590 and various versions of the firmware, however the sawtooth packet loss pattern is there during a period when I had a draytek plugged in as well as a Zen 7530 loaner

If it was the Fritzbox, then the sawtooth pattern would most likely reset on reboot and that's not happening. The other issue is those wild ping latency swings at times when the connection is quiet

That's why I suspect the issue isn't CPE and much more likely to be inside Openreach. I'm on an ancient ECI cabinet and given the issues they're known to have I suspect the issue is the local exhange's node controller, but BTOR simply won't talk to anyone about their issues

Matters aren't helped by wet cables, but my VDSL has been fairly solid for several months, vs previous link instability

I'm aware of Fritz ICMP flood control, but it seems to have been somewhat fixed in more recent versions

Have you seen the entry regarding sawtooth patterns on this website ?

Why does my graph have a sawtooth effect?

One unusual but common effect is shown in the sample graph below. The repeated rising edge with a sharp drop off gives rise to the waveform name of sawtooth.

BQM sawtooth pattern showing jitter affecting max latency caused by some routers

A graph of this sort appears to be common with users of the Sky LLU broadband service. It does not appear to represent an actual problem with the service, or affect any applications running across the connection. This is just one of those odd situations that a particularly router appears to cause.

Not the same sawtooth. In the OP's case, the sawtooth is of increasing packet loss over a 6 hour period that suddenly reduces and climbs again.

The FAQ example is a sawtooth pattern in the latency over a much, much shorter time period.

Definitely not the same thing.

Graph

So this is with a Fritz on a Zen FTTC line. Terrible performance, esp. with VOIP... not happy.

Edited by davidmatthewson (Fri 24-Jan-25 13:47:21)

This is... interesting... Zen have been telling me there are no similar cases

What part of the UK are you in (SE here, LSLEA) and what kind of cabinet are you connected to? (Huawei or ECI?)

If there are more people with similar issues it adds to the corpus that there's a very real issue they need to deal with properly, instead of just brushing people off

We are in SE UK - East Sussex - The Cab is ~50m from the router and identies as 'Broadcom' but maybe that's in the Exchnage rather than cab?

So, potentially a regional issue

The identfied chipset is what's in the DSLAM, not the controller (exchange)

Broadcom modems are Huawei cabinets. ECI uses Infineon - that rules out exchange node manager issues as they're different for each maker and pushes the issue further upstream

==

Thanks for the info - just to be clear, are you suggesting any issue is in the Cab, Exchange or Router - sorry to be dense!

Edited by davidmatthewson (Fri 24-Jan-25 19:05:17)

In reply to a post by Stoatwblr:

This is... interesting... Zen have been telling me there are no similar cases

What part of the UK are you in (SE here, LSLEA) and what kind of cabinet are you connected to? (Huawei or ECI?)

If there are more people with similar issues it adds to the corpus that there's a very real issue they need to deal with properly, instead of just brushing people off

Sadly that is Zen's standard response nowadays.

Yes and extremely disappointing

Unfortunately this is what happens when accountants take over companies

I suspect I'll be moving on if it's not resolved. A&A beckons

==

Given the lack of commonality, it can't be routers or cabinets and is unlikely to be the exchange controller (they're different for Huawel and ECI even if both are in the same exchange - as is the case in Leatherhead due to a replacement program being underway when Boris's edict on Huawei came down and stopped it along with the FTTH rollout)

It's possible that the issue is the regional node in BT's backhaul to Zen and it would be interesting to see if the sawtooth shows up other areas of the UK or only in the Southeast

I think I have a similar problems of dropouts, 2 or 3 times a day lasting a minute or so (often once during early hours) since the last week of 2024. I am on IDNet SOGEA but I believe ZEN provide the backhaul. Of course when I speak to IDNet support the system no problems are found. I have done all the usual internal checks trying other modem, router and cables, and I am convinced the problem is on the other side of the master socket. I am in Cheltenham (GL51).

My Broadband Ping

OK, I've done some more playing and replaced the Fritz with a Draytek 2865 - and it is rock solid. The saw tooth dropped packets display vanishes totally. So some Fritz artifact perhaps? [see pics]
https://pixeldrain.com/u/ZAw7dW4w
https://pixeldrain.com/u/gHq99gfY
I'm staying with the 2865 for now as I find the Fritz a bit too basic for what we need - esp. VPNs and no SNMP.

D

Edited by davidmatthewson (Wed 29-Jan-25 09:51:25)

The pattern in the graph you just posted is completely unlike what I'm seeing and looks like a different issue

I have no doubt you're seeing "something" with the Fritz. The question is what it is

Given past experience as an ISP owner and banging heads against Telcos, I'm fairly sure the issue is inside Openreach and that solving it will be a long and involved process

When persistent line/connection problems exist, a good ISP willing to push the issue is like gold dust. There's a huge need to collate data to look for common faults on one cabinet/concentrator.

In one instance the data I'd collated on faults reported/experienced by MY customers enabled the telco to verify that field contractors were "farming" faulty cables by simply moving customers onto different pairs as the complaints came in - once the telco was aware of the scale of the issue they replaced half the street feeders in one town and all the issues went away (This also made me public enemy number one as far as those contractors were concerned because their steady income was slashed by 80%)

In this instance the problem is considerably further upstream and bears some resemblence to what I saw with overloaded Frame Relay connections in the 1990s