Kitz Forum

Broadband Related => ADSL Issues => Topic started by: Weaver on December 26, 2015, 01:29:35 AM

Title: Problem with speed testers again - presumed packet loss
Post by: Weaver on December 26, 2015, 01:29:35 AM
See
     http://forum.kitz.co.uk/index.php/topic,15742.msg295256.html#msg295256

This issue has arisen once more. I've just changed to three brand new modems (DLink DSL-320B-Z1) from Andrews & Arnold, and things are suddenly extremely slow - as measured by two different speed testers.

I'm assuming from the experience of the old thread that packet loss is happening again?

The new modems are configured specially for me with the xDSL mode selection option set to mode='ADSL2-only' (or whatever) as opposed to 'auto', defaulting presumably to ADSL2+, but we don't know that for sure. This is supposed to be the only difference in their config.

Does anyone have any thoughts?

Title: Re: Problem with speed testers again
Post by: Weaver on December 26, 2015, 01:55:59 AM
See A & A continuous quality monitoring server pic for one of the lines :

   https://www.dropbox.com/s/yigu9mgeokml6su/image.png?dl=0 (https://www.dropbox.com/s/yigu9mgeokml6su/image.png)
Title: Re: Problem with speed testers again
Post by: Weaver on December 26, 2015, 02:00:59 AM
The pic shows "dripping blood" the red streaks coming down from the top, an indication of packet loss.

See http://aa.net.uk/kb-broadband-cqm.html (http://aa.net.uk/kb-broadband-cqm.html)
for a description.
Title: Re: Problem with speed testers again - presumed packet loss
Post by: aesmith on December 26, 2015, 11:25:23 AM
I must admit that the term "packet loss" is one that sort of sets my teeth on edge (same with VoIP pronounced as a word).  Reason being that drops are a fact of life in networks except in very specific cases.

However this might be one of those cases.  As I understand it A&A are polling with LCP, being roughly speaking the PPP equivalent of ICMP.   Are these LCP packets (frames?) prioritised over other traffic in both directions?   If so then loss of a poll should only occur if it's dropped within BT, or if there's an error or a drop between the DSLAM and your router.   Or do the replies come from your modem? 
Title: Re: Problem with speed testers again - presumed packet loss
Post by: Weaver on December 26, 2015, 01:46:35 PM
> Are these LCP packets (frames?) prioritised over other traffic in both directions?   
Has to be so, yes. As the PPP LCP echo requests are sent by a special server in a Firebox 6000.

> If so then loss of a poll should only occur if it's dropped within BT, or if there's an error or a drop between the DSLAM and your router.

Absolutely.

> Or do the replies come from your modem?

The replies do come from the EU's modem.
Title: Re: Problem with speed testers again - presumed packet loss
Post by: Weaver on December 26, 2015, 01:48:13 PM
See http://aa.net.uk/kb-broadband-cqm.html (http://aa.net.uk/kb-broadband-cqm.html)
Title: Re: Problem with speed testers again - presumed packet loss
Post by: aesmith on December 26, 2015, 05:11:37 PM
> Are these LCP packets (frames?) prioritised over other traffic in both directions?   
Has to be so, yes. As the PPP LCP echo requests are sent by a special server in a Firebox 6000.
And the replies also prioritised by your equipment?
Title: Re: Problem with speed testers again - presumed packet loss
Post by: WWWombat on December 26, 2015, 06:45:16 PM
I'm wondering about what this graph shows in your individual case, Weaver.

You work with a bonded connection, don't you? Where does the L2TP connection terminate? In the modems? Or in a separate piece of equipment?

What I'm really wondering is whether you have one graph that shows the status of the bonded connection, or 3 graphs that show the components.

I'm mostly concerned about the purple sections - the ones that signify total outage.

If the graph represents the bonded connection, then purple sections likely mean all 3 modems are down. If the graph represents a single link, then the purple sections indicate a lack of sync in one modem - and we could do to see the other two graphs. Both ways, they suggest a lack of basic stability, and this is what should be sought first.
Title: Re: Problem with speed testers again - presumed packet loss
Post by: Weaver on December 26, 2015, 08:46:27 PM
> You work with a bonded connection, don't you?
Indeed, three IP-bonded lines. (Not ML-PPP)

> Where does the L2TP connection terminate?
It goes from the BTW BRAS to a server at Andrews & Arnold.

> In the modems? Or in a separate piece of equipment?

Misunderstanding. See previous reply. the modems speak PPP and don't know anything about L2TP or IP.

>  or 3 graphs that show the components

There are always three graphs, one per line. I just picked the one of interest. (When all is well, of course their output is very similar.) One pic shows red streaks coming down from the top of the graph, which represents loss of PPP LCP 'pings'.

In the second pic, the purple section shows that line is down. There was a fault in the Broadford exchange and the modem was just being made to cycle, gaining sync, then going down and round again.

* It seems the link I gave out was wrong, it points to an image folder in the Dropbox website, not to an individual image, and the second image is intended for discussion in another thread. Very confusing! It just shows I don't know how to use uploaded images properly.
Title: Re: Problem with speed testers again - presumed packet loss
Post by: burakkucat on December 26, 2015, 09:13:50 PM
I think the attached image, below, is what was intended to be shown via the "Dropbox" link . . .  :-\

(b*cat is rather confused and feels that a quick nap is overdue.  ???  )
Title: Re: Problem with speed testers again - presumed packet loss
Post by: Weaver on December 26, 2015, 09:26:43 PM
Saved by Burakkucat!