In reply to a post by Stevenage_Neil:

For obsessive don't you mean "have OCD"

Oh that's good, now we have a doctor on this thread.

indeed,

the way I noticed was 2 things.

1 - when I migrated from BT retail, lost 1-2mbit of throughput (cant remember exact amount) even tho I was on the same sync speed and IP profile.
2 - when I started the ipv6 trial, the profiles arent automated as far as I am aware and as a result I had a profile that was too high (in terms of plusnet) so in affect only had the BTw ip profile limiting meaning when I started the ipv6 trial my speeds went up.

In reply to a post by Oliver341:

In reply to a post by mixt:

I think I've got more important things to worry about in my life than the speed of my broadband connection. And I also think some others haven't.

And you post that on a forum full of people who are interested/obsessive about broadband?

It would appear that way, wouldn't it? His rant about going to Ofcom if his sync speed drops 3 mbits is just laughable. Especially on a connection that is 75-80mbits anyway. That's not even a 10% drop. Further more, the speed "as advertised" (by the ISP) is up to 76mbits. Which is what he has. Not to mention other factors (cross talk) that may lower sync speed further over time.

Out with any of this, Zen who I'm looking at moving to next month clearly state in their small print regarding garunteed speed that, for a 80/20 service, they will provide a minimum speed of 40mbits or 70% of line speed, for at most 90% of the busiest 3 hours of the day (which ever is less). So if I'm signing up to that, I accept that I may not get my full throughput speed 100% of the time.

I just don't get some people on here when it comes to speed tests. It's like they pay peanuts and expect to run a speed test every 10 minutes with it to come back at full speed 100% of the time. These people will be doing the same thing on a gigabit FTTP link if they had one, and would then complain if their speed test reported a 1% drop in what it should be. Seriously, give me a break.

Ok, in this instance, it looks to be something to do with Plusnet traffic shaping, and the reasons for that have been explained by several people elsewhere on this thread (including myself). But the OP knew Plusnet traffic shaped before signing up, did he not?

</end of rant>

Edited by deleted (Mon 24-Mar-14 08:17:13)

Say what you like about adslmax but one thing he can do is start a thread that runs and runs.

Don't knock the PHD! Some people just don't advertise.

Copied from the Plusnet forum.

Are you only measuring the speeds using speedtests? This unfortunately is slightly flawed. We've spent a lot of time ensuring we get the profiling correct and there are a lot of reasons for that.

First, the reason we have the profile is two-fold, first the spec of the product we buy from BTW means that we are supposed to apply a profile on our side to ensure that the we don't send a customer more data than they can receive. In the event of a DDoS attack for example someone could be sent several hundred Mbps or even several Gbps of traffic, without a profile our side that would head down the BTW network and hit the BRAS which in the most severe event could have a knock on effect to 10s of thousands of people on that BRAS.

BTW send us regular updates when speed profiles change so we can keep our side up to date, so knowing we have to do something we want to be clever and smart about it. We take the BTW profile and round down to the nearest 100kbps or 200kbps depending on whether you're on ADSL or fibre.

The second reason and one of the big differences Plusnet has over most other ISPs is the way we do our traffic management. The Plusnet profile is a scheduler or a shaper. That means that when the amount of traffic being downloaded is greater than or equal to the profile rate we will buffer the traffic. Without the profile or if the profile was set above the BTW profile the traffic would hit the BTW profile limit which is a policer. The difference is that this would cause packet loss or retransmissions.

The trouble with speedtests is that they will see the few ms that the download is in the buffer as a slow down in the download rate which it doesn't take into account.

Try out a download or look at the utilisation in the interface. Whenever we've tested this out we've always seen that the actual throughput coming down the wire is faster than the speedtest reports and in testing in the labs this was the behaviour we saw.

The third reason for the Plusnet profile is that we can use it to prioritise one type of traffic over another when you max out the line. If you're watching a video and making a VoIP call and download via P2P all at the same time you shouldn't see problems with the video or the VoIP but instead see the P2P max out at whatever bandwidth is going spare rather than try and swamp the other 2 protocols.

Are you actually trying to say that TCP overheads mean that there is more data travelling over the connection than if you just measure how long to transfer a 100 MB file?

In the labs what we saw was that when you started the download, the download speed at the start would try and go faster than your line could support but the way that we manage the traffic means that any data that can't go down the line immediately goes in the buffer and gets a minor delay added to it.

A 100MB file should take virtually the same amount of time to download on a Plusnet line as any other ISP of the same sync speed but the buffering might mean slightly less data travels over a Plusnet connection to transfer said file because you might see a couple less packets if there's any retransmits on the line without the buffering.

It's just a slight behaviour difference but as I say this is what we saw with a lot of testing on the values to use with our profiling to get the best experience. In some cases speedtest results gave slightly lower results but the actual data transferred and the time taken was no different.

Looking on the network the shape of the graph was kind of like this:

/\________
|..............\

And when we moved the the buffer we saw.
___________
|...............|

Same amount of data, same amount of time but a different pattern.

I've always wondered how the speedtest sees the buffer emptying at the end.

There's some good stuff over on Google about improving TCP transfers:

http://googlecode.blogspot.co.uk/2012/01/lets-make-t...

The way we do it might reduce the number of round trips because it could reduce the number of retransmissions.

I don't get it

But Dave...

Have you read my post above?

I can see a clear 3Mbps difference in speeds, between the fake 20M profile and the proper 77.4M profile (both PN profiles)... with an 80Mbps sync speed (and 77.4 BT profile) in both cases.

The *only* difference is the two PN profile values.

Does that mean your 77.4M profile is deliberately de-rated, so that the traffic management prioritisation (of, say, voip) can function? That I lose 3Mbps all the time, just in case I max out the line occasionally?

(To those who think this search for 3 Mbps is obsessive, you're wrong. I don't notice nor care that it is missing, but I do care that PN's system might be wrong. It seems to be a systemic problem).

Edit:
Last time I had the 20M profile was a while ago, when my account placed restrictions on the throughput some of the day. If you want to try some experiments with my line, I'm game to try that, with results measured on the SamKnows box.

Edited by deleted (Tue 25-Mar-14 01:50:19)