FTTC Speed Tests

I have noticed discussions in various threads regarding online speed test accuracies.

I don't know for sure which are the most accurate, but the most consistent ones I use are speedtest.net & mybroadbandspeed.co.uk

They both give reasonably similar results & appear to be consistent with their reporting.
e.g. when my connection is poor, I get poor results.

I personally prefer speedtest.net as it logs your test results over longer periods than mybroadbandspeed.co.uk, & these results can be easily exported to a csv file for really long term monitoring etc. (also recording the IP address in use at the time of the test, so showing re-syncs etc.).
Registration with speedtest.net is not required for these basic & useful facilities.

The TBB speed tester can, on occasions, appear to give poor results, even when a user's connection is otherwise "O.K."

The usual BT performance/speed tester ALWAYS produces slightly more pessimistic results, but apart from the occasions when it experiences obvious "problems", is also quite consistent.
It does take much longer to conduct & can only be conducted once every hour, so not so good for instantly monitoring what happens if you should choose to reboot/re-sync the modem.

To me, consistency is the most important aspect & switching between speed testers just to get a better result doesn't assist at all.

Always using the same speed tester, regardless of whether results are ALWAYS pessimistic or ALWAYS optimistic, at least demonstrates a definite pattern to your connection's real-life performance (subject to any contention / traffic management for various types of traffic etc).

So, in addition to sticking with the same speed tester, I believe that speed tests should always be conducted at the same time of day/evening/night to truly monitor a connection's performance over as prolonged a period as possible.

Again, for consistency reasons, timing the download of a fixed sized file at regular times of day appears to confirm connection performance patterns.

One of BT's URLs for just this purpose can be found at this link:-
BT Test Files

The largest file is only 15MB (15,675,392 bytes on disk), which was probably ideal for much slower ADSL1 connections, but it IS a constant to use even over FTTC connections.

e.g. For my own connection from when it was "problem-free" (speed tests & ability to download at up to 33 Mb), I could usually download the 15MB file in 3 seconds.
It currently takes 5 seconds (pretty much at any time of day or night).

When my connection was variously performing even worse than it is now, it could take up to 14 seconds.

I accept that the "accuracy" of the timings is a bit loose (whole seconds only), but again, over a prolonged period a pattern emerges, & a sudden difference really does highlight "problems".

I would be curious to see timings for downloading the 15 Mb file from other users, especially those whoose speed tests are around 33 Mb.

The above is purely my own opinion, so please feel free to agree/dsagree accordingly.


Paul.

P.S. I have no idea what the test file (largedownload.me) contains, but suspect it is just typical & consistent "data", specifically designed for testing purposes.

Edited by deleted (Wed 21-Dec-11 09:21:10)

Those old BT files bring back memories from the old 500kbps days!

From memory the data was data was created in such a way that compression application that claimed to "increase bandwidth" would have little effect so the real capability could be tested.

I have just tried the Largedownload.me and it gave me 4Mbps - rather slow. My connection should run at 36Mbps but has a few problems but not that bad!

Remember: Delete the file after it has been downloaded.

Hi MHC,

In reply to a post by MHC:

Those old BT files bring back memories from the old 500kbps days!

From memory the data was data was created in such a way that compression application that claimed to "increase bandwidth" would have little effect so the real capability could be tested.

I have just tried the Largedownload.me and it gave me 4Mbps - rather slow. My connection should run at 36Mbps but has a few problems but not that bad!

Remember: Delete the file after it has been downloaded.

I get:-

Downloaded: 14.9MB in 5 sec

Transfer Rate: 2.98MB/Sec

Regardless of whether I delete it, or simply overwrite it when asked.
Overwriting it is the quickest method for me.

Did you mean it took 4 seconds, or did you mean MBps rather than Mbps?

Paul.

Generally two test methodologies exist

1. Average over long period - us
2. Sampling, with rejection of some high and low values to arrive at a consensus - ookla

This means we are more likely to give a low result even if there is a short stall or period of congestions, whereas ookla ones will reject this, thus giving something more akin to learning what your headline speed is. For example you can pull your ethernet out during an ookla test and as long as its plugged back in before the end get a decent result.

Another difference is the protocols used, by utilising HTTP ookla generally avoids traffic management, we have a custom protocol, that providers could spot but many don't and lump our test with the P2P stuff.

Once the tests are averaged over large numbers of tests, we do tend to broadly agree with all the testers, i.e. individual circumstances become negliable in a big sample.

It was around 502 kBps in MS terminology! or 4Mbps in real data transfer termnology.

Yes the units are right!

It was always a case of deleting - MS would create Largedownload(1).me, Largedownload(2).me ... I had a colleague who complained of low disk space and we found over 100 of the files on her PC.

Thanks for that explanation Andrew,

To me, it is a very clear explanation, that should hopefully address many of the queries I have seen recently (not just in the TBB forum).

Paul.

In reply to a post by MHC:

It was around 502 kBps in MS terminology! or 4Mbps in real data transfer termnology.

Yes the units are right!

Hmmm. Not looking too healthy at all.

Is it currently just a problem with that particular link/test file?


Paul.

They may be hosted on a 100Meg only link, and posting the URL may have lead to a burst in use.

In reply to a post by MrSaffron:

They may be hosted on a 100Meg only link, and posting the URL may have lead to a burst in use.

Possibly, but I repeatedly ran my own tests around that time, still maintaining the same 2.98MB/Sec.

As I've posted elsewhere, I am puzzled that the BT official test, the Ookla-based test sites and the TBB test site differ so much.

BT and Ookla (e.g. speediest.net) consistently report me at 37/38Mbps down, 8/9Mbps up; TBB consistently says 18/19Mbps and 8.5Mpbs respectively. I wish there were a way of knowing which numbers best related to real-world benefits.

TBB tester often fails to keep up. Nothing new.

Just for curiosity, what do you get from this link (15MB test file), & how does it compare against your other speed tests?

http://www.btopenworld.com/speedtest

Paul.

Edited by deleted (Wed 21-Dec-11 11:44:36)

About 4.5s to download that, which is a bit too short for reliable bandwidth measurements I think. It works out around 28/29Mbps.

I agree, the file is realistically too small for FTTC speeds, but it's the only one I could find that appears to be specifically geared to testing throughput speeds.

If you know of any others, please shout out accordingly.

Well, using TBB's own download test files, on the 100MB file the throughput comes out at about 23Mbps, which is near 30% greater than their own test reports.

Different transfer protocol, you will be using HTTP for the file download and providers can handle different categories of traffic in different ways.

In the same way that people complain when RTSP (flash streaming) often can be slower than speedtest results.

Thanks for the clarification. What's your own take on these measured differences? I see your comment about the different measurement design philosophies and personally wouldn't bother if the difference was in the order of a few percent, but essentially TBB is reporting somewhere in the region of half the speed of most other speed test sites.

And does your ISP employ traffic management for P2P?

To really comment one needs to see a tbbmeter plot, with indication which is our test, and which is the other test. Or another usage graph tool if you want.

We have no Windows O/S in the household, alas, so cannot use TBBmeter. If you happen to think of a suitable Mac one, I'd be happy to give it a go.

As for traffic shaping/management, perhaps you experts can tell me whether my ISP, BT Infinity, employs stuff like that?

Not a recommendation, but what I could quickly find

http://www.skoobysoft.com/utilities/utilities.html - Surplus Meter

No, I have a problem which; I cannot resolve, BT second line cannot resolve, 2x technician visits cannot resolve, router replacement has not resolved ... Modem stats look OK and I will get round to running your batch files! Currently waiting for a Router firrmware update which "might" solve it. My feeling is that my problem is in the cabinet!

In reply to a post by MHC:

My feeling is that my problem is in the cabinet!

You, and the public sector unions.

I get the same results with the TBB speedtest showing my connection alot slower than other speedtests.

Thanks for sharing. That's almost exactly the picture I see in all my tests on the Ookla platform.

The MLAB test suite seems pretty authoritative: it's from a consortium featuring Google amongst other distinguished players, and offers tools to measure bandwidth, traffic shaping/throttling/blocking and a bunch of other diagnostics I've not seen anywhere else. I'm well impressed.

Their test gives me 3.1MB/s (~25Mbps). It also reports that BT Infinity appears to rate-limit my uploads (specifically, BitTorrent and port 6881 uploads achieved only 969Kbps), but no download rate limiting was detected.

From this discussion it would appear that TBB reports are minima, Ookla and others are maxima, and real-world figures somewhere in between. Interesting, methinks. I shall focus on the elaborately documented and peer-reviewed MLAB tests in future.

Thanks� I believe this is more a consumption monitoring tool than a systematic bandwidth measuring tool. Please see my comments re: the MLAB test suite.

My line, with 38714 profile, just gave a 37.5Mbps result on speedtest.net.

With the 15MB file, I tried using "wget" on my Linux box, which fetched the file using the HTTP protocol. This resulted in a range from 3.5 to 4.1 seconds.

1) 3.5s = 4.47MBps = 35.8 Mbps (where M = 1,000,000)
2) 4.1s = 3.8MBps = 30.6Mbps.

Those are raw data speeds, without adjusting for the overheads of HTTP or TCP.

Which of the speedtesters present their answer in raw end-end data terms, and which one adjust it to make allowances (possibly guessed) for the various overheads?

In reply to a post by WWWombat:

My line, with 38714 profile, just gave a 37.5Mbps result on speedtest.net.

That is probably spurious - is it repeatable? I normally get just over 35 on speedtest.net, but have on occasion got over 40, which is clearly impossible.

Edited by kasg (Wed 21-Dec-11 16:58:42)

Do you still have the same problem if your PC is connected directly to the modem (by-passing the router completely)?

In reply to a post by RobertoS:

In reply to a post by MHC:

My feeling is that my problem is in the cabinet!

You, and the public sector unions.

I sort of work in the public sector, so I got it, & very topical too

In reply to a post by bruiser1:

Well, using TBB's own download test files, on the 100MB file the throughput comes out at about 23Mbps, which is near 30% greater than their own test reports.

I tried that 100 Mb file at around noon today.

It took 1 minute 37 seconds, with a transfer rate of 1.03 MB/Sec. i.e. roughly one third of my usual download speed of around 3 MB/Sec.

Traffic managed by my ISP?

In reply to a post by kasg:

In reply to a post by WWWombat:

My line, with 38714 profile, just gave a 37.5Mbps result on speedtest.net.

That is probably spurious - is it repeatable? I normally get just over 35 on speedtest.net, but have on occasion got over 40, which is clearly impossible.

It is indeed repeatable, depending on the PPP session.

Basically, I can get a PPP session (established either because I connected it via the router GUI, or because I rebooted the modem, or unplugged the phone line for a bit) and I then get a stable download speed. That stable speed will be either around 35Mbps or around 37.5Mbps.

I've reported this to Plusnet on this thread.

I haven't found a foolproof way of getting a "fast" session, but it seems to have happened more often when I've caused a modem resync (by unplugging the phone line) and just waiting for the router to re-establish itself a few minutes later; I've not managed very much when I've done a manual disconnect & connect using the router GUI.

Very interesting, thanks, I might try some experiments myself.

Use this link below for BT FTTC test file! It a 500 MB large file for up to 40Mbps FTTC

http://82.20.174.2/speedtest/download/500MB?sc_1317060825_639

In reply to a post by adslmax:

Use this link below for BT FTTC test file! It a 500 MB large file for up to 40Mbps FTTC

http://82.20.174.2/speedtest/download/500MB?sc_1317060825_639

In reference to 82.20.174.2:

These servers are for intended for internal use only. However, if you need help with your Virgin Media broadband, why not visit our new forum, just click here.

Yep it a real true speed for fibre optic. I know it belong to virgin media but can test it to ensure you get the right speed! Virgin tech guy gave to me ages ago when I was complaint to him of all speedtest is never correct to my high speed 100 Meg before.

Eh? It is just a big file stored on a server on the VM network, just as many other files are.

It would be better to find a file on the BT Retail network if the person is using Infinity, i.e. avoid peering issues and might also bypass traffic management which can be IP, port, protocol based