Kitz ADSL Broadband Information
adsl spacer  
Support this site
Home Broadband ISPs Tech Routers Wiki Forum
 
     
   Compare ISP   Rate your ISP
   Glossary   Glossary
 
Please login or register.

Login with username, password and session length
Advanced search  

News:

Pages: 1 ... 5 6 [7] 8 9 10

Author Topic: Speed issues with AAISP  (Read 21618 times)

jdigz777

  • Just arrived
  • *
  • Posts: 6
Re: Speed issues with AAISP
« Reply #90 on: July 05, 2018, 04:09:53 AM »

Will Uno eventually have all it's users moved to the new TT network?

All of uno's customers are on the New TT network. I believe for now I am the only customer who has been moved back to the old network.
Logged

PhilipD

  • Reg Member
  • ***
  • Posts: 591
Re: Speed issues with AAISP
« Reply #91 on: July 05, 2018, 07:49:07 AM »

Hi

Currently my Uno Sync is 79999 @ 1.1dB SNR and am seeing a cap of 68 mbits. But if I use the Uno speedtester it indicates 72 mbits but before May the Uno speedtester was giving me 77 mbits download.

The reason I haven't reported to Uno is my attainable sync is 75 mbit as my SNR has dropped since the winter so if I re-sync it will be moot point.

The reason the Uno speed test provides a higher reading is the result is multiplied by 1.06, this is a default adjustment made by the open source code they use to cater for overheads (A&A use the same open source tester code and likely do the same thing), so that speed test isn't your data throughput, but overall throughput. Thinkbroadband (and other speedtesters like speedtest.net) report the actual data throughput which is more meaningful.  Therefore if you take your 72Mbps from the Uno test and divide by 1.06 you get, wait for it, a spooky answer of 67.9Mbps.

The evidence is definitely there isn't it, TalkTalk is limiting the maximum throughput.

Regards

Phil
Logged

PhilipD

  • Reg Member
  • ***
  • Posts: 591
Re: Speed issues with AAISP
« Reply #92 on: July 05, 2018, 07:56:34 AM »

Hi

Will Uno eventually have all it's users moved to the new TT network?

I'm with Uno and had no notification of a move to the new network and only knew of it from A&A status page.  Around the same time as the AA status page indicated users were being moved over that is when I noticed my speed tests slowed and I then appeared to have a throughput ceiling lower than the sync rate should provide.  So I must be on the new network.  It should be noted that prior to the move I had no issues with congestion or latency, and the new network probably increased overall pings and latency by a few ms.

Regards

Phil
Logged

PhilipD

  • Reg Member
  • ***
  • Posts: 591
Re: Speed issues with AAISP
« Reply #93 on: July 05, 2018, 08:41:35 AM »

Hi Andrew

Your speed test at http://speedtest2.aa.net.uk/ which uses the open source test https://github.com/adolfintel/speedtest has the default compensation factor applied of 1.06, so it reports (or tries to) the total throughput rather than actual data throughput, so it will always appear to be faster than other tests.

If you look in the Javascript code, in speedtest_worker.js and search for overheadCompensationFactor, setting this to 1 will give a truer speed test and comparable to most others.  Currently at the moment for the TalkTalk ceiling we are seeing, the A&A speed test comes out as ~74Mbps (as does Uno's), whereas it is really ~68Mbps.  It's almost as though TalkTalk have tuned their throughput settings to get 74Mbps on that test, which being inflated due to trying report throughput including overheads, has then resulted in everyone see a ceiling of 68Mbps.

Regards

Phil

« Last Edit: July 05, 2018, 08:43:58 AM by PhilipD »
Logged

andrew-AAISP

  • ISP Rep
  • Member
  • *
  • Posts: 41
    • aa.net.uk
Re: Speed issues with AAISP
« Reply #94 on: July 05, 2018, 09:22:48 AM »

I could do with a few (A&A) customers trying to max out their line, and then letting me know their login so I can review our CQM graphs.

...Our graphs records all sorts of data, including the throughput and when/if the line gets full - ie transfers traffic at the rate of the line. this is shown by red pixels at the base of the graph.

So far I have tested two lines, one line does not get the red pixels and one does - hence I need some more examples.

I suggest doing all of the following downloads simultaneously for about 15 minutes:

a) Download a torrent or two of a Debian iso (or similar) -
  https://cdimage.debian.org/debian-cd/current/amd64/bt-dvd/debian-9.4.0-amd64-DVD-1.iso.torrent

b) Downloads of 1G files from these two places:
    http://ipv4.download.thinkbroadband.com/1GB.zip
    http://speedtest2.aa.net.uk/1GB.zip

c) Run iperf3 as well, a few times against our server (if possible)

d) run a Netflix speed test a few times:
    www.fast.com

This should be done from a wired connection if possible, the speed testers will show low speeds due to all the other traffic but that's expected.

What we're looking for is the red line as seen on the bottom of this example graph: https://imagebin.ca/v/47cIr93UTAvP

Report back via support@aa.net.uk with subject: TalkTalk speed tests
Logged
A&A

andrew-AAISP

  • ISP Rep
  • Member
  • *
  • Posts: 41
    • aa.net.uk
Re: Speed issues with AAISP
« Reply #95 on: July 05, 2018, 09:29:47 AM »

Hi Andrew

Your speed test at http://speedtest2.aa.net.uk/ which uses the open source test https://github.com/adolfintel/speedtest has the default compensation factor applied of 1.06, so it reports (or tries to) the total throughput rather than actual data throughput, so it will always appear to be faster than other tests.

If you look in the Javascript code, in speedtest_worker.js and search for overheadCompensationFactor, setting this to 1 will give a truer speed test and comparable to most others.  Currently at the moment for the TalkTalk ceiling we are seeing, the A&A speed test comes out as ~74Mbps (as does Uno's), whereas it is really ~68Mbps.  It's almost as though TalkTalk have tuned their throughput settings to get 74Mbps on that test, which being inflated due to trying report throughput including overheads, has then resulted in everyone see a ceiling of 68Mbps.

Regards

Phil

I've set this to 1
Logged
A&A

jelv

  • Helpful
  • Kitizen
  • *
  • Posts: 2054
Re: Speed issues with AAISP
« Reply #96 on: July 05, 2018, 09:37:32 AM »

I could do with a few (A&A) customers trying to max out their line, and then letting me know their login so I can review our CQM graphs.

...
Do you only need this for TT lines?
Logged
Broadband and Line rental: Zen Unlimited Fibre 2, Mobile: Vodaphone
Router: Fritz!Box 7530

jelv

  • Helpful
  • Kitizen
  • *
  • Posts: 2054
Re: Speed issues with AAISP
« Reply #97 on: July 05, 2018, 09:46:07 AM »

I've set this to 1

Just run another test on a BT 80/20 sync line. I have the line rate set at 100%.
Logged
Broadband and Line rental: Zen Unlimited Fibre 2, Mobile: Vodaphone
Router: Fritz!Box 7530

PhilipD

  • Reg Member
  • ***
  • Posts: 591
Re: Speed issues with AAISP
« Reply #98 on: July 05, 2018, 09:56:30 AM »

Hi Andrew

I've just had a look and that hasn't changed in your speed test.  This is because your test is pulling in the minimised file from http://speedtest2.aa.net.uk/speedtest_worker.min.js and not the more human readable one.  I would suggest the easiest option is in the Html page change the line "w = new Worker('speedtest_worker.min.js')"  to use the none minimised file (speedtest_worker.js) with the setting now set as 1 (Uno are using the non minimised version), and you then have an easy way to roll back to how it was if required, or perhaps set up a second HTML page using the non minimised version of that script so we can compare and contrast.

Regards

Phil









Logged

andrew-AAISP

  • ISP Rep
  • Member
  • *
  • Posts: 41
    • aa.net.uk
Re: Speed issues with AAISP
« Reply #99 on: July 05, 2018, 10:20:42 AM »

Do you only need this for TT lines?

Happy for BT lines too - it makes a good comparison.
Logged
A&A

andrew-AAISP

  • ISP Rep
  • Member
  • *
  • Posts: 41
    • aa.net.uk
Re: Speed issues with AAISP
« Reply #100 on: July 05, 2018, 10:24:46 AM »

Hi Andrew

I've just had a look and that hasn't changed in your speed test.  This is because your test is pulling in the minimised file from http://speedtest2.aa.net.uk/speedtest_worker.min.js and not the more human readable one.  I would suggest the easiest option is in the Html page change the line "w = new Worker('speedtest_worker.min.js')"  to use the none minimised file (speedtest_worker.js) with the setting now set as 1 (Uno are using the non minimised version), and you then have an easy way to roll back to how it was if required, or perhaps set up a second HTML page using the non minimised version of that script so we can compare and contrast.

I've done that now!
Logged
A&A

PhilipD

  • Reg Member
  • ***
  • Posts: 591
Re: Speed issues with AAISP
« Reply #101 on: July 05, 2018, 10:47:59 AM »

Hi Andrew

Just testing that, and indeed it's changed now and reporting lower, however it's gone a bit too low. I've not got time to look at the code to see what it is doing, but currently my speed test shows as 64.98 with the factor at 1 (i.e. no change), I was expecting something around 69.  If we take 64.98 * 1.06 it equals 68.88 which is what I would expect from that change.  I'm guessing perhaps a bug in the test code is applying that correction factor twice and perhaps that is the bug, not the actual factor itself, therefore a factor of ~1.03 would correct that more or less.  I'm guessing at the moment so I don't suggest changing it unless you want to, perhaps you would prefer to revert back to how it was?

All other speedtests are currently coming in around 68-69Mbps very consistently, the only exceptions are those based on the fdossena open source tester, I've tried Uno, yours, and even their own test page one at http://speedtest.fdossena.com/, they all report higher where dividing by 1.06 matches other testers. 

Edit:  Had a quick look at the Javascript and can't see this being applied twice, however the help documentation has:

Code: [Select]
overheadCompensationFactor: compensation for HTTP and network overhead. Default value assumes typical MTUs used over the Internet. You might want to change this if you're using this in your internal network with different MTUs, or if you're using IPv6 instead of IPv4.
Default: 1.06 probably a decent estimate for all overhead. This was measured empirically by comparing the measured speed and the speed reported by my the network adapter.

1048576/925000: old default value. This is probably too high.
1.0513: HTTP+TCP+IPv6+ETH, over the Internet (empirically tested, not calculated)
1.0369: Alternative value for HTTP+TCP+IPv4+ETH, over the Internet (empirically tested, not calculated)
1.081: Yet another alternative value for over the Internet (empirically tested, not calculated)
1514 / 1460: TCP+IPv4+ETH, ignoring HTTP overhead
1514 / 1440: TCP+IPv6+ETH, ignoring HTTP overhead
1: ignore overheads. This measures the speed at which you actually download and upload files rather than the raw connection speed

So 1 seems to be the true download speed, all the others are trying to estimate the overheads, and it can be seen from the notes that most aren't calculated, but arrived at by some guess work and observations.  Interesting 1.0369 is listed as an alternative, which is what I suggested above would seem to give rise to something in the ball park of other testers.

Seems like this isn't a precise science where this tester is concerned, I wonder what adjustments are made by other speed testers?

There is a newer version of this tester code which has various bug fixes as well.

Regards

Phil

« Last Edit: July 05, 2018, 11:16:28 AM by PhilipD »
Logged

andrew-AAISP

  • ISP Rep
  • Member
  • *
  • Posts: 41
    • aa.net.uk
Re: Speed issues with AAISP
« Reply #102 on: July 05, 2018, 11:26:55 AM »

Yes, a bit tricky isn't it!

I've set it to 1.0369 for now, and have upgraded it to v4.5.4

Also, I was wondering if it was worth having a longer test, so there is now a 5 minute download test at
  http://speedtest2.aa.net.uk/graph5.html
That is worth trying to see if speed changed as time goes on.

I've not tested this on a DSL line though.

Logged
A&A

ktz392837

  • Reg Member
  • ***
  • Posts: 559
Re: Speed issues with AAISP
« Reply #103 on: July 05, 2018, 11:30:57 AM »

I think care is needed now and no knee jerk changes made.  You have changed the speed tester so that it is reporting less speed than what it was so if the TT network improves you can't compares figures.  You may have even changed the tester so it is under reporting speeds now so it is still broken and any data collected with it before the change is now null and void.  Urrgh.
Logged

PhilipD

  • Reg Member
  • ***
  • Posts: 591
Re: Speed issues with AAISP
« Reply #104 on: July 05, 2018, 11:33:48 AM »

Hi Andrew

Yes, a bit tricky isn't it!

I've set it to 1.0369 for now, and have upgraded it to v4.5.4

Also, I was wondering if it was worth having a longer test, so there is now a 5 minute download test at
  http://speedtest2.aa.net.uk/graph5.html
That is worth trying to see if speed changed as time goes on.

I've not tested this on a DSL line though.



Hi Andrew

You are working quick, yes can see you've changed it to 1.0369, this is giving me around 70Mbps, which is much closer to all the other testers at around 69Mbps at the minute when I've just done a quick test around them. So certainly in the ball park now and seems more indicative of elsewhere.

The longer test flat-lines at the 70 mark as well.

Hopefully others can confirm this test is now matching others.  I suppose the acid test is if you see tests that are not physically possible from an 80/20 sync, i.e. too high, which would indicate the compensation is slightly over compensating perhaps, but for now that seems to match other tests a lot better.

Regards

Phil



Logged
Pages: 1 ... 5 6 [7] 8 9 10