OK, I've used both tools, the download manager never gets above 4900KB/s (so ~ 38Mbps), and I can see that in the tbmeter while it's running.

Interestingly tbmeter does show one peak of 53Nbps, but I guess that must be a statistical anomoly as it only every happened the once and right in the middle of the transfer.

The tbmeter monitored download was nothing like as consistent with the download as the download manager was though, the rate fluctuated up and down widely from 16 to 40 Mbps over the course of the download.

I re ran the download manager test, this time setting the tbmeter frequency to 1/4 second and I didn't get the same flat profile as before, it jumps about a lot but does list spikes of up to 48 Mbps, though never more than two of these in a row, so the sustained transfer rate is still sub 40 Mbps.

Is it possible that tbmeter is just misreporting the stats, or is the throughput really that eratic.

No other monitoring tool has ever reported anything over 41 Mbps.

I just need to be clear on what I'm seeing here before I put all this evidence together and report to my ISP.

In reply to a post by davestubbs:

The tbmeter monitored download was nothing like as consistent with the download as the download manager was though

I've seen the same.
With the TBB multi-thread (5 threads) 25MB download, compared to a 20MB (5 thread) download, from the TBB site, but using a download manager to do the 20MB file, I see typical download speeds of 12.5Mbps from the TBB meter download (can vary maybe 2Mbps either side, but I haven't yet seen a TBBmeter multi-thread download go higher than around 15Mbps). Yet with a TBB sourced 20MB file, split into 5 threads using a download manager and downloaded seconds later, I get around 32Mbps.
This is repeatable. Time and again I get at least 2-3 times the download speed from a download manager, as from TBBmeter (I'm glad I'm not the only one to be seeing this) and the graph in TBBmeter reflects this difference (so it's not simple a case of the download manager giving a false download speed, or the TBBmeter giving a false download result in the opposite direction).

The 20MB download (using the download manager) appears on TBBmeter as a larger spike with less than half the duration (even accounting for the 5MB file size difference) as the 25MB download initiated by TBBmeter itself.

To equalize things I set TBBmeter to 4 threads (to get it to select a 20MB file for its multi-thread download test).
TBBmeter downloaded in 16.3 seconds (a download speed of 10.292Mbps).
My download manager did the 20MB TBB test file, again with 4 threads, in under half the time and at more than double the speed.

It seems it's better to go by the download manager than TBBmeter (at least for checking download speeds - single thread or otherwise).
I have no explanation for the difference (over to Andrew for more investigation work - not that he's anything better to do ).

If it were possible to imbed images into this forum I could post a screen-shot of the TBBmeter graph.

One assumes the TBBmeter is getting its test file from the same server as the TBB 20MB download test file.

To test the actual capability of my line I downloaded the TBB 512MB file, using a download manager set to 20 threads, and got a download speed of a little over 73Mbps (so nothing wrong with the line nor the ability of the TBB servers to upload at high speeds).

Edited by adebov (Sat 05-May-12 20:17:58)

The fact it takes 20 threads to fill the link is very telling, as same code on other providers has of late shown speeds of over 200 Mbps from same sources.

I would not be surprised if my multithreading is not as effecient as the download managers.

The weight of evidence is leaning towards something I've seen before and that is at times of congestion BT wholesale services tend to throttle single thread type downloads.

Alas while full time around here now, there are so many jobs to get done.

In reply to a post by MrSaffron:

The weight of evidence is leaning towards something I've seen before and that is at times of congestion BT wholesale services tend to throttle single thread type downloads.

Which is pretty much what I've been saying all along and I seriously doubt if any amount of phone calls to BT Retail's help line in India will get around a BT Wholesale single thread throttle
It's not a massive issue as most large downloads can be performed by a download manager, but it's been particularly bad since BT rolled out 80/20. Up until a few weeks ago I was getting much higher single thread speeds and only the occasional sign of throttling (now the throttle seems to kick in at 5pm weekdays and all weekend).
I hope it doesn't get any worse (as most of us are locked into stupidly long contracts - the sign of a poor ISP is needing to force customers to stay, rather than assuring customers stay because the service is so good).

In reply to a post by MrSaffron:

Alas while full time around here now, there are so many jobs to get done.

I know the feeling!

In reply to a post by adebov:

Which is pretty much what I've been saying all along and I seriously doubt if any amount of phone calls to BT Retail's help line in India will get around a BT Wholesale single thread throttle

Are you saying the throttling is in the WBC network then? Surely this should be hard to determine given every packet is encapsulated in PPPoE frames? (or PPPoA on ADSL) ??

I hope it doesn't get any worse (as most of us are locked into stupidly long contracts - the sign of a poor ISP is needing to force customers to stay, rather than assuring customers stay because the service is so good).

If it is WBC rather than GEA implemented - then people on TalkTalk and/or Sky shouldn't see this - but people on ADSL24 / IDNet / PlusNet should.

Any more data yet?

In reply to a post by jchamier:

Are you saying the throttling is in the WBC network then? Surely this should be hard to determine given every packet is encapsulated in PPPoE frames? (or PPPoA on ADSL) ??

I don't know. Probably depends on whether BT is able to tell the destination IP (and thus; that's it's one of theirs) and implement the throttle that way.
It's either that or I'm imagining it and I actually have 75Mbps speeds on single download threads (and TBBmeter is lying in reporting anywhere between 2.5 to 15Mbps per thread, along with various speedtesters and download managers).
Something MUST be going on (affecting the download speed of each thread). The line is able to go at full speed, if I get enough threads going. It's also able to get full speed, via speedtesters, providing it's during the normal working day.
I'm struggling to think of any explanation, other than a peak time single thread speed cap, which would cause speeds to suddenly plummet at around 5pm weekdays (also rubbish speeds right through the weekend, from 5pm Friday) whilst full speed is still attainable via multi-thread downloads.
I've yet to see a router/modem/software/pc/anti-virus fault which only presents itself during evenings & weekends.

In reply to a post by jchamier:

If it is WBC rather than GEA implemented - then people on TalkTalk and/or Sky shouldn't see this - but people on ADSL24 / IDNet / PlusNet should.

There are reports, on these forums, from the odd few customers on FTTC seeing poor speeds.

The thing is there are so few reports, at this stage, that it's equally possible BT have a bad-boy throttle.
It's certainly not widespread (which also made me think of exchange or even cabinet throttles, if capacity is about to be exceeded).

For what it's I don't see any issues on my download. TBB is usually a tad lower than other testers but that's normal. The upload result is still wrong but that's a different thread. I'm with IDNet and so far have no reason to suspect that anyone anywhere is throttling me.

In reply to a post by adebov:

I don't know. Probably depends on whether BT is able to tell the destination IP (and thus; that's it's one of theirs) and implement the throttle that way.

Then I would hazard a guess that any throttle is on the BTretail / Infinity part of the connection, not the OR street level or WBC backhaul parts ??

There are reports, on these forums, from the odd few customers on FTTC seeing poor speeds.

But are they all Infinity or WBC users, or are they TalkTalk and Sky users too?

Or is it that we don't have enough data yet?

In reply to a post by jchamier:

Or is it that we don't have enough data yet?

Nail on the head!
Too few Sky FTTC customers for any useful information there (same may apply to Talk Talk, but given TT's appalling reputation with ADSL I guess any TT customer will to overjoyed with whatever fibre speed they can get).

Andrue is on the same exchange as I (although he's in a neighbouring town, their cabinets are connected, by fibre, to our exchange so he'll be using the same WBC back-haul, out of the exchange) and he only reports the odd issue with upload speed (but that could be more to do with the TBB speedtester, rather than an actual upload speed issue). So that kind of rules out exchange back-haul as a reason to throttle.

I've no idea how many people are activated on our cabinet, but we're right on the edge of town with very poor ADSL speeds so I'd have thought take-up would be higher than on cabinets nearer the exchange.
So cabinet back-haul is still a possibility and we know OR set a default limit for the cabinet link (to the exchange) with a view to having an easy future upgrade path should cabinet take-up be higher than expected.

I don't know whether IDnet are simply reselling a BT retail product, or using their own transit links / pipes (or whatever they're called this week).

So we're still looking at cabinet congestion (I don't know if BTw have a mechanism for throttling connections at cabinet level (I rather suspect they do - I mean; you can set QOS on many domestic routers so it seems unthinkable a similar facility won't be included in the Huawei equipment in the cabs).
Or BTw/Retail throttling (I doubt this as there are enough people, on Infinity, reporting no speed issues to take the view this isn't happening).
Then we have so called bad-boy throttling (from a service which is supposed to be unlimited with the FUP removed last April).

I haven't received an e-mail, from BT, suggesting I'm over their FUP so will have restricted speeds this month (any 'decent' ISP would surely warn users if they're downloading too much).
In any case; I don't think I've been anywhere near the old 300GB FUP so that (coupled with the fact there's not even supposed to be an FUP) should rule out a bad-boy pipe or bad-boy throttle.

So still undecided and nowhere near enough data to form a conclusion.

Edited by adebov (Sun 06-May-12 16:10:21)

I am not aware of any ISP re-selling BT Retail. IDNet certainly do not and never have. They use WBC for FTTC just like all ISPs except TT and Sky. IDNet have their own MSILs, the WBC equivalent of 20CN BT Centrals.

BT Wholesale have no direct control over the cabinet DSLAMs. Unlike the DSLAMs/MSANs in the exchanges, where BT Wholesale have their own like any LLU supplier. The cabinet DSLAMs are Openreach. BTW will be able to set the QOS settings as specified in SIN 498, sections 2.1.5 and 2.2.7. But that isn't throttling as such. As I read it, it is prioritising VOIP and suchlike.