In reply to a post by Tam:

it is network management that is the cause.

It may well be network management - but network management by whom, and for what purpose? Not all network management comes down to the plusnet traffic shaping system (which is designed to place various limits on your connection, restricting individual protocol usage at various times of day).

The fact that shifting gateway can correct the problem does indeed suggest the problem does not come down to the traffic shaping system, but that the problem rests elsewhere.

I'd suggest the overall gateway is congested, thus the overall gateway network management is having to apply shaping to certain protocols to ensure that there is sufficient bandwidth is available to things that need it.

That could certainly be true - some leg/router is congested, and has started to employ congestion reduction techniques. That isn't shaping, in the same sense as shaping associated with your Plusnet account.

If it was congestion without network management then *everything* would have been slow,

Not necessarily. Weighted fair queuing techniques can affect long slow transfer more than short, fast ones.

Similarly, congestion detection avoidance techniques, such as RED (random early detection) can deliberately trigger early packet loss to slow down random TCP transfers; to the end user, it looks like some transfers are running slowly.

WRED can do the same, but with weighted traffic classes. The end-user experience will look very similar to traffic shaping, but is actually just plain network management and congestion prevention, although slightly more intelligent.

The kicker is that these impacts could be happening within the BTO and BTW domains, as well as within Plusnet. For one of the Plusnet users complaining about signs of congestion recently, the cause came down to congestion within certain BTW routes in Leeds.

Simply out of interest, reading this thread and checking the gateway I'm on, how can I be on a gateway that doesn't seem to exist on the 'Collated Users Online graph' displayImage.php

In reference to usertools.plus.net/@gateway/:

Which gateway You are currently connected to gateway ptn-ag02.

This is located in Telehouse North.

There is no ptn-ag02 ?

I think it's labelled as ptn-ag2 (no 0) on the graph.

Yes.

In reply to a post by WWWombat:

It may well be network management - but network management by whom, and for what purpose? Not all network management comes down to the plusnet traffic shaping system (which is designed to place various limits on your connection, restricting individual protocol usage at various times of day).

I think all the people reporting the problem are on the current products, which do not have any time of day parameters. Pure traffic type prioritisation.

I was disconnected around 8am along with a number of other users according to the useage graphs (looks on the graph as if there were at least 4 significant instances of drops in user connections since yesterday). I've been struggling to get off the 'bng' gateways where my latency was around 35ms. The speedtest graph also looked odd with initial surges and drops before settling down at an even 36-38Mbps.

I've finally got onto 'ptn-ag03' and my latency has dropped to 24ms and that initial 'bounce' on the speed test graph has been replaced with a steady rise and drop to a straight line still at 36-38Mbps.

In reply to a post by MrSaffron:

It could just be a bug in a bit of software too.

It is *slightly* reminiscent of an issue I had with IDNet which they eventually traced to a mis-configured LNS. I don't think this is the same thing though because in that case single threaded downloads had a more obvious cap whereas what I see when impacted on PN is single threaded speeds all over the shop.

To me it looks like congestion and my current theory is that something is throttling on a 'per stream' basis and multi-threading works around this because it uses multiple streams. So I am inclined to suspect the PN QoS system but why it should feel the need to kick in is a whole other question. My connection was fine last night and as far as I can tell from my TBBQM I'm still on the same gateway that I was on a week ago.

Edited by Andrue (Thu 22-Jan-15 15:54:55)

I don't believe they are segmented.

I suspect the issue will be somewhere between PN and the WMBC Shared handovers at Faraday/Colindale. This would make sense as you can reconnect on the same gateway, but end up on a different handover point or host link to that specific gateway.

The bng gateways are a load of rubbish. I always get poor performance on them.

Somehow I got onto one the other day without noticing, no re-sync), and after a couple of days of normal web pages taking 3-4 seconds to load instead of the normal under one second checked the latency and gateway. Latency was up at 20ms from a normal 12-14ms.

Speed tests however were unaffected.

After two attempts I got onto a non-bgn and everything returned to normal.

I find a disconnection of around a minute usually results in a gateway change. A few seconds doesn't.

I found that the BNGs gave me higher latency (6-9ms to bbc.co.uk) compared to the AGs (3-4ms to bbc.co.uk).

I've recently found though that sometimes I could land on a BNG and have the 3-4ms again. I came to the conclusion that it wasn't the gateways, but the links to them.

I'm really splitting hairs though when talking about a few ms - it makes no difference to me at the end of the day.