Kitz Forum

Announcements => Site & Forum Discussion => Topic started by: UncleUB on August 22, 2016, 06:39:14 PM

Title: Speedtesters
Post by: UncleUB on August 22, 2016, 06:39:14 PM
Just noticed on the main site pages that under the list of speedsters bbmax is no longer in use.It says on the site that the licence has expired. :)

http://www.speedtest.bbmax.co.uk
Title: Re: Speedtesters
Post by: GigabitEthernet on August 22, 2016, 08:46:40 PM
Been like that for some months I think.

I find dslreports the most accurate. I find (and have always found) speedtest.net to give impossible numbers.
Title: Re: Speedtesters
Post by: Black Sheep on August 22, 2016, 09:20:19 PM
I find speedtest.net to be the best.
Title: Re: Speedtesters
Post by: kitz on August 22, 2016, 09:56:13 PM
Thanks - removed.
Title: Re: Speedtesters
Post by: GigabitEthernet on August 23, 2016, 12:06:53 AM
I find speedtest.net to be the best.

I find it awful. It always says my speed is at least 2Mb faster than it should be.
Title: Re: Speedtesters
Post by: Black Sheep on August 23, 2016, 07:38:32 AM
I use it every day, on perhaps 4/5 different EU's and different BB products. I have been doing so for many years now.

I don't work to decimal points, haven't really got the time  ;), but I will show the EU the connection speed of their router via our WHOOSH system, and explain that on a wired connection they should be getting a percentage of this (roughly 94%), but always explain that being anal about it will drive them to an early grave.

I genuinely can't recall anyone questioning the speedtest.net result ............ they're more than happy to have gone from poor speeds to very fast speeds, and the results are always there or thereabouts the 90%+ area.
Title: Re: Speedtesters
Post by: Weaver on August 23, 2016, 07:53:18 AM
The speedof.me one seems to be rather higher than the others for me. I'm quite a fan of the thinkbroadband one. It would be better if the various speedtesters made it easier to choose between IPv6 and IPv4.
Title: Re: Speedtesters
Post by: roseway on August 23, 2016, 08:00:56 AM
Speedof.me always returns results which exceed the maximum possible for me. :) The truth is of course that none of them are accurately representative of real life, and should only be taken as approximate indicators, in the way that BS uses them.
Title: Re: Speedtesters
Post by: j0hn on August 23, 2016, 08:35:17 AM
I mainly use speedtest.net but it does occasionally give an impossible result, i.e higher than my current sync
I do have to manually select a server myself though, as it always determines Carlisle to be the best and for whatever reason it's always a good 5-10Mbps slower than others
Title: Re: Speedtesters
Post by: GigabitEthernet on August 23, 2016, 12:35:41 PM
DSLReports I find to be really good. Especially the bufferbloat test.
Title: Re: Speedtesters
Post by: jelv on August 23, 2016, 01:20:14 PM
The thinkbroadband one is my favourite because of the single thread and 6 thread comparison. That often reveals congestion when others tests don't.
Title: Re: Speedtesters
Post by: Weaver on August 23, 2016, 09:54:42 PM
I find some of them dubious, possibly because of my distance from London? and have to pick servers carefully for the ookla one for example. The dslreports one has only "London" as a relevant choice amongst a lot of remote international ones iirc? The dslreports one is a minefield
 of info though, and I like the fact that a couple offer IPv6 testing.
Title: Re: Speedtesters
Post by: NEXUS2345 on August 23, 2016, 11:45:21 PM
One that I tend to use is one provided by a server host I use. proof.ovh.net (http://proof.ovh.net)
I think it is pretty good, not really had any issues with it over reporting, and OVH have a very efficient network between the UK and their datacentres in France.
Title: Re: Speedtesters
Post by: kitz on August 24, 2016, 11:06:47 AM
I find some of them dubious, possibly because of my distance from London? and have to pick servers carefully for the ookla one for example. The dslreports one has only "London" as a relevant choice amongst a lot of remote international ones iirc? The dslreports one is a minefield
 of info though, and I like the fact that a couple offer IPv6 testing.

I would still choose a London server if I were you.  AAISP's gateways are in London (Docklands?) so choosing anywhere else will just add on additional time, latency and possible points of congestion outside the ISP control.
Some people assume they should select the server nearest their home location, which isn't true.  If possible choose a speedtest location near to your ISP's gateway, otherwise your traffic could go all the way down the country and then backup again before reaching the speedtest server.

See tracerts below to show the difference between the different server locations....  and how using the Lancashire server increases latency as traffic has to come halfway back up the country again for no good or useful reason.

Speedtest.net @ Lancashire server

Code: [Select]
Tracing route to ookla.bt.lancashire.co.uk [72.52.4.91]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2    12 ms    12 ms    12 ms  lo0.10.central10.pcl-bng01.plus.net [195.166.130.138]
  3    12 ms    12 ms    12 ms  irb.10.PCL-CR02.plus.net [84.93.249.82]
  4    13 ms    13 ms    12 ms  ae2.pcl-cr01.plus.net [195.166.129.6]
  5    12 ms    12 ms    12 ms  ae1.ptw-cr01.plus.net [195.166.129.0]
  6    14 ms    14 ms    14 ms  blackhole.prolexic.com [195.66.224.31]
  7    17 ms    16 ms    17 ms  unknown.prolexic.com [72.52.60.192]
  8    15 ms    15 ms    15 ms  unknown.prolexic.com [72.52.60.199]
  9    16 ms    17 ms    16 ms  unknown.prolexic.com [72.52.4.91]

Trace complete.


Speedtest.net @ Maidenhead server

Code: [Select]
Tracing route to speedtest-net-1.mh.speedtest.uk.net [213.229.119.169]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2    12 ms    12 ms    12 ms  lo0.10.central10.pcl-bng01.plus.net [195.166.130.138]
  3    12 ms    12 ms    12 ms  irb.10.PCL-CR01.plus.net [84.93.249.81]
  4    12 ms    12 ms    12 ms  ae1.ptw-cr01.plus.net [195.166.129.0]
  5    13 ms    12 ms    12 ms  the-linx.as29550.net [195.66.224.223]
  6    14 ms    13 ms    13 ms  ae1-cr0.rdg.as29550.net [91.186.5.250]
  7    14 ms    14 ms    14 ms  b.3.magic-hex.as29550.net [92.48.95.62]
  8    14 ms    14 ms    13 ms  mirror.as29550.net [213.229.119.169]

Trace complete.
Title: Re: Speedtesters
Post by: d2d4j on August 24, 2016, 11:43:37 AM
Hi

I hope you don't mind, but plus.net have used our speedtest server recently

Here's the link to the test if your interested

Many thanks

John

http://www.speedguide.net/speedtest/results_img.php?test=4573622
Title: Re: Speedtesters
Post by: Weaver on August 24, 2016, 03:31:55 PM
@kitz I did some experiments a while ago on many of the ookla servers dotted around England. I didn't make the mistake of assuming that a North of England server would be the fastest.
For me the London ones, although they are, as you rightly say, are expected to be the fastest, came out with slightly lower readings than ones around the M4 corridor to the west of London. Iirc Newbury or Maidenhead was the fastest. Now this could I suppose possibly have changed if I did my tests before AA did expensive network upgrades in Docklands at the start of the year.

However, I told myself at the time that perhaps the London ones were simply very busy, very popular, either heavily loaded themselves or on pipes that carried a lot of traffic related or unrelated.
Title: Re: Speedtesters
Post by: Weaver on August 24, 2016, 04:11:12 PM
If you have a server 100 mi north of London, and my ping has to go down to London and then back up northwards (for one half of the round trip), then that will be an effective extra round trip time of 0.7 ms if you assume an (effective) refractive index of 1.48 for single mode optic fibre. (From stackexchange for physics of sm fibres and the properties of materials.) This is nothing much compared to my ping times to ookla servers reported as >40 ms. (Does this mean the duration of half or of the whole round trip time?)

So geography is, for me anyway, a lot less important than extra hop count or busy servers or links.
Title: Re: Speedtesters
Post by: Weaver on August 24, 2016, 04:17:15 PM
Ookla speedtest iPad app, a few tests just after 15:40 Wed:

47 ms: 6.31 / 0.80 and 6.32 / 0.83 Mbps - Maidenhead, XILO
44 ms: 6.43 / 0.82 and 6.37 / 0.80 Mbps - Newbury, Vodafone
43 ms: 6.48 / 0.83 and 6.42 / 0.81 Mbps - London, Vodafone
42 ms: 6.56 / 0.86 and 6.47 / 0.67 Mbps - London, Coreix
Title: Re: Speedtesters
Post by: Weaver on August 24, 2016, 04:36:53 PM
Northern servers

63 ms: 4.72 / 0.82 - Aberdeen, Internet For Business Ltd
55 ms: 6.40 / 0.80 - Carlisle, Gemini Group
53 ms: 6.51 / 0.85 - Aberdeen, Converged Communication Solutions

And one very odd, noisy, result from the furthest south:

60 ms: 6.51 / 0.70 Mbps, and 50 ms: 6.19 / 0.69 and 6.50 / 0.69 - BT Lancashire Services
Title: Re: Speedtesters
Post by: GigabitEthernet on August 24, 2016, 04:41:28 PM
My ping to Canterbury is apparently 1ms! Is that where TalkTalk's gateways are?

(http://www.speedtest.net/result/5576853890.png) (http://www.speedtest.net/my-result/5576853890)

London is around 7-8ms.
Title: Re: Speedtesters
Post by: Weaver on August 24, 2016, 04:48:35 PM
If I were really organised, I could work out what the IP address of those ookla servers is, by observing the active TCP or UDP connections reported by the os. And then I could traceroute to them just to see what the network topology is like at least.

That's not going to really give an answer to GigabitEthernet's question though, not if we want real geography.
Title: Re: Speedtesters
Post by: Weaver on August 24, 2016, 04:53:03 PM
It shows you anyway that you do have to be really careful, as the best result for download is 1.37 times the worst. And some servers, though not all, have some variation from one test to another. I'm also assuming, hopefully, that there was no inbound or outbound activity on my own internet connection at the time of the tests, as that would really stuff things up, but I suspect I would see that in the pings' times?
Title: Re: Speedtesters
Post by: GigabitEthernet on August 24, 2016, 05:01:09 PM
That 1ms result is clearly wrong. If I perform a ping on the IP/hostname of the second hop of any tracert, the best ping I can get is 5ms. That result is impossible.
Title: Re: Speedtesters
Post by: kitz on August 24, 2016, 10:57:42 PM
If I were really organised, I could work out what the IP address of those ookla servers is,
That's not going to really give an answer to GigabitEthernet's question though, not if we want real geography.

No need.  Theres a list here (https://www.speedtest.net/speedtest-servers.php).   
Thats the list I used to be able to do my tracerts earlier (http://forum.kitz.co.uk/index.php/topic,18337.msg331465.html#msg331465).  Just search for the location name and it will give you the host name, and a trace should automatically give you the IP.


Quote
That's not going to really give an answer to GigabitEthernet's question though, not if we want real geography.

1 ms is impossible.   Its usually a minimum of 6ms to the RAS or core entry point (depending on who the ISP is or if its WBC).


Code: [Select]
<server url="http://speedtest.thinkdedicated.com/speedtest/upload.aspx" lat="51.2750" lon="1.0870" name="Canterbury" country="Great Britain" cc="GB" sponsor="Cloud Space UK" id="1531" host="speedtest.thinkdedicated.com:8080"/>
It even gives the longitude and latitude of the locataion :)



Code: [Select]
Tracing route to www.thinkdedicated.com [91.239.125.30]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2    12 ms    12 ms    12 ms  lo0.10.central10.pcl-bng01.plus.net [195.166.130.138]
  3    12 ms    16 ms    14 ms  irb.10.PCL-CR02.plus.net [84.93.249.82]
  4    13 ms    13 ms    13 ms  195.99.126.98
  5    14 ms    13 ms    13 ms  core3-te0-19-0-21.faraday.ukcore.bt.net [109.159.249.159]
  6    14 ms    13 ms    13 ms  213.137.183.34
  7    13 ms    13 ms    13 ms  ldn-b3-link.telia.net [213.248.67.97]
  8    13 ms    13 ms    13 ms  ldn-bb3-link.telia.net [62.115.117.6]
  9    13 ms    13 ms    13 ms  ldn-b3-link.telia.net [62.115.137.201]
 10    13 ms    13 ms    13 ms  THINK-SYSTE.ear1.London1.Level3.net [195.50.124.74]
 11    13 ms    13 ms    13 ms  91-239-125-30.thinkdedicated.com [91.239.125.30]

Quote
Iirc Newbury or Maidenhead was the fastest.
Yes I always use Maidenhead for the same reason.   A lot of Plusnets gateways are also at Docklands, although they do have some central london.

Title: Re: Speedtesters
Post by: kitz on August 24, 2016, 11:29:22 PM
Hi

I hope you don't mind, but plus.net have used our speedtest server recently

Here's the link to the test if your interested

Many thanks

John

http://www.speedguide.net/speedtest/results_img.php?test=4573622

Added :)
Title: Re: Speedtesters
Post by: Weaver on August 25, 2016, 03:44:09 AM
For some unknown reason, I just get a blank page on my iPad in Safari when I follow the link
     https://www.speedtest.net/speedtest-servers.php

I tried in another iOS web browser, iCab, too. Although possibly they share too much common code
Title: Re: Speedtesters
Post by: jelv on August 25, 2016, 08:22:30 AM
When looking for the 'closest' server you must take in to account where your connection leaves the ISPs network and connects to the internet. So for example on Plusnet a user in Inverness is over 800 miles away from the speedtest.net server hosted by Internet for Business in Aberdeen.

(This was certainly the case on their old network - it may not be true on the new - I don't know)
Title: Re: Speedtesters
Post by: kitz on August 25, 2016, 12:05:12 PM
For some unknown reason, I just get a blank page on my iPad in Safari when I follow the link
     https://www.speedtest.net/speedtest-servers.php

I tried in another iOS web browser, iCab, too. Although possibly they share too much common code

Its just a php webpage that outputs all the servers info in xml format.   There's no style sheet to make it look pretty and its just plain text output.
A quick google shows that iCab on iOS doesn't support xml parsing. [it requires a Microsoft XML parser which is not available on the iPad.]
It looks like Safari on iOS doesnt have an inbuilt xml parser either.

Title: Re: Speedtesters
Post by: Weaver on August 25, 2016, 12:20:02 PM
Ah. They could do with some minimal CSS for the XML then. At least it's not me going mad. ;D