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:

Author Topic: Underlying reason for odd differing speedtest results on the upstream side  (Read 3305 times)

tommy45

  • Reg Member
  • ***
  • Posts: 563

This issue of differing results  with upstream throughput as reported by speed test sites ,

 BTW-PT 16-17mbps max,
 Think-broadband 18.40-18.60 (normally )
 Speedtest.net, 18.70-19.00 but does vary depending on the sever selected even within the uk

And several other speed-testers  tend to give results which are  50% or less the  of actual throughput  i can achieve ,

Today  tbb is giving 12.-14mbps for the upstream,  is this an issue with tbb or one of the peering partners used ? as said above  usual results are 18+mbps today following a gateway change from a dodgy bng gw to a pcl  gw,, or is it down to my isp ?
 Although this isn't a big issue, it does bug me, never had this issue with other isp's and i would like to know possible  or most likely causes
Logged

PhilipD

  • Reg Member
  • ***
  • Posts: 395

Hi

On TTB I always get exactly 18.71 upload (sync'd at 20) and usually a bit more variable on other speed tests but always 18+.  Down speeds are usually ~74 but often at peak times due to what appears to be the usual BT congestion can drop and judder around 30 or 40Mbps, however even in these cases upload speed is still 18+, I presume due to the asynchronous nature of xDSL, i.e. more is being downloaded than uploaded.

Your speeds could vary for several reasons perhaps:

1) Something on your computer or network is using data and uploading at the time of the speed test
2) Your IP stack on the computer, maybe try resetting to defaults if Windows especially if you have tried tweaking the settings in the past.  Windows 7 and above rarely need tweaking anymore, and often when you do it favours one speed test but not another.
3) Router issue, perhaps a QoS setting or CPU too busy to handle the traffic at full speed
4) Line issue, errors on the line causing packet loss and requiring a resend
5) Virus/firewall software on your computer that may be inspecting packets

then, perhaps the more likely

6) Throttling by the ISP or congestion, this will be the case if a gateway change resolves it.  Are the slow downs during peak times?

Years ago with PlusNet they constantly had this issue with ADSL where the gateways would become unbalanced with more people on one than the others, and often they would drop thousands of connections to force them to re-balance.  I'm sure this was only a problem because they ran the network with zero spare capacity.  Not sure this now applies, but changing gateways and seeing a performance improvement takes me back to such a time.

Regards

Phil

 
Logged

tommy45

  • Reg Member
  • ***
  • Posts: 563

My OS is win XPSP3 ,the upstream throughput on tbb speed tester, did return to normal later on, and yes some evenings my throughput does slow during the peak time 7pm-12am usually it's only the downstream affected, but there has been a few occasions where both upstream & downstream affected, As for Gateway swapping, that doesn't always fix or improve  the problem, but the issue is usually only during  the busiest times

So  i doubt it has much to do with my PC the software running on it, or the router , as if any of those ,i wouldn't be seeing the issues only at certain times of the day and not every day,  I would think it be also unlikely though not impossible that it isn't down to errors, and there is next to no packet loss  on my tbb graphs, and the majority of what there is, Is down to the networking used  from ISP to tbb, from certain Gateways  or  endpoints , as apparently both can change each time you restart  your PPPOE session, I have also observed another anomaly  and that is some Gateway's provide higher throughput levels  than others for some reason , as well as one of two GW's having a 6-7ms increase in latency. with some others quite a few , adding 2-3ms  to what you should normally see



Logged

PhilipD

  • Reg Member
  • ***
  • Posts: 395

Hi

Always worth trying from another PC if possible.  Background tasks that start and stop can have a random effect, and if your PC hardware is older then interrupts being raised can slow down Ethernet.  Also the speed test requires CPU cycles to operate, if I'm maxing out my quad core PC with video encoding the speed tests are lowered considerably.

Like I suggested however, most like the ISP.

Regards

Phil
Logged

Chrysalis

  • Content Team
  • Addicted Kitizen
  • *
  • Posts: 5720

Indeed, my spare pc cannot get full speed. Due to CPU saturation in chrome but can with not much excess power in IE or firefox.  (for whatever reason chrome requires tons of cpu for flash stuff).  It maxes out a dual core amd.
Logged
Sky Fiber Pro - Billion 8800NL bridge & PFSense BOX running PFSense 2.4 - ECI Cab - LINE STATISTICS CLICK HERE

tommy45

  • Reg Member
  • ***
  • Posts: 563

As i run Everest ultimate which monitors the cpu  utilization  i can safely say it's not  my cpu, No IMO it's more like the server how it's set up for upload testing, or and the path taken to that server,

Speedtest.net different servers give different results more so on the upstream , my cpu utilization whilst conducting  tests  is max 50% but usually lower , more so whilst uploading
Here is an example  

« Last Edit: June 10, 2014, 05:46:06 PM by tommy45 »
Logged

Chrysalis

  • Content Team
  • Addicted Kitizen
  • *
  • Posts: 5720

yeah I know.
Logged
Sky Fiber Pro - Billion 8800NL bridge & PFSense BOX running PFSense 2.4 - ECI Cab - LINE STATISTICS CLICK HERE

boost

  • Reg Member
  • ***
  • Posts: 763

All of these speedtests are inherently unreliable, IMO.

Ideally, some sort of test just inside BT's network using a looow overheads protocol would be as close as you can get to consistent, measurable, throughput.

Assuming your OGHPs aren't too congested I spose :D
Logged

tommy45

  • Reg Member
  • ***
  • Posts: 563

All of these speedtests are inherently unreliable, IMO.

Ideally, some sort of test just inside BT's network using a looow overheads protocol would be as close as you can get to consistent, measurable, throughput.

Assuming your OGHPs aren't too congested I spose :D

I never assumed any speed test to be reliable, But the differing results between them are consistent for some reason or other, I never had this issue when on ADSL , So it seems to be an issue with the upstream being nearly 20mbps, But some others appear to get better results than i do, the speedtest  @ http://www.speedtest.bbmax.co.uk/ always gives a lower  result than my real  upstream throughput, but i have seen other FTTC customers get more accurate upstream results, which could indicate it's not the speed test host, IMO i think it is down to the ISP, in some way
Logged

Chrysalis

  • Content Team
  • Addicted Kitizen
  • *
  • Posts: 5720

All of these speedtests are inherently unreliable, IMO.

Ideally, some sort of test just inside BT's network using a looow overheads protocol would be as close as you can get to consistent, measurable, throughput.

Assuming your OGHPs aren't too congested I spose :D

Odd I dont tend to find them unreliable.
Logged
Sky Fiber Pro - Billion 8800NL bridge & PFSense BOX running PFSense 2.4 - ECI Cab - LINE STATISTICS CLICK HERE

boost

  • Reg Member
  • ***
  • Posts: 763

All of these speedtests are inherently unreliable, IMO.

Ideally, some sort of test just inside BT's network using a looow overheads protocol would be as close as you can get to consistent, measurable, throughput.

Assuming your OGHPs aren't too congested I spose :D

Odd I dont tend to find them unreliable.

I'm sure they are mostly reliable if not very reliable but

1. We can't control the path there or back
2. We can't control load along the path
3. We can't control server load
4. We can't control the number of threads launched to 'saturate' our connection nor the logic used to invoke those threads

There's prolly more. Lots more :P
Logged

Chrysalis

  • Content Team
  • Addicted Kitizen
  • *
  • Posts: 5720

thats the idea, its not just a line speed test, but an actual internet speed test, so yes the peering/transit capability of the isp is also been tested.

the speedtest server load's dont saturate.

in my mind the only part I consider unreliable is that it is multithreaded meaning test results can be higher than real world usage if single threaded speeds cant match.

the tbb speedtester does single threaded tests as well as multithreaded so that is a very good speedtest.

you have to remember the internet isnt just one fat path (this I think you understand anyway), there is different routes traffic can take, so the speedtest result should be reliable for that route/destination but of course not representative of all of the internet, for this reason on speedtest.net people can choose different speedtest servers anywhere in the world.

Using a speedtest server inside the isp or bt wholesale will be less representative.
« Last Edit: June 12, 2014, 08:53:55 AM by Chrysalis »
Logged
Sky Fiber Pro - Billion 8800NL bridge & PFSense BOX running PFSense 2.4 - ECI Cab - LINE STATISTICS CLICK HERE

boost

  • Reg Member
  • ***
  • Posts: 763

Yep, you're right :)
Logged