[TelstraClear] Is the Telstraclear speedtest reliable?

Forums › Vodafone New Zealand (including Vodafone, WxC, Farmside) › [TelstraClear] Is the Telstraclear speedtest reliable?

OldFart

36 posts

Geek

#114775 2-Mar-2013 23:00

Something seems weird about the TC speedtest. If I run a speed test on the Telstraclear site I get around 100 Mbps (I have a warpspeed connection).
But if I do the same thing on other speed test sites I get nothing like 100 Mbps.
Call me suspicious, but am I really getting "warp speed" anywhere other than on the TC site?

From the Consumer test:

From speedtest.net (Auckland)

From the Telstraclear speed test

Talkiet

4427 posts

Uber Geek

Trusted

#773765 2-Mar-2013 23:48

It is absolutely believable that a speedtest server on the same network as you are would give much better results than others, even if they are in the same city or country.

As for warp speed, I'm unaware if the TCL residential broadband service has any performance guarantees, especially to sites beyond their own network - but I _stongly_ suspect it doesn't.

Cheers - N

Please note all comments are the product of my own brain and don't necessarily represent the position or opinions of my employer, previous employers, colleagues, friends or pets.

DoomlordVekk

129 posts

Master Geek

Trusted

#773789 3-Mar-2013 01:37

Ian, its the design & the dimensioning of the TCL(now VF) network that says "All things being equal, you should be able to achieve your plan speed across the TCL network".

Its also in the way the speedtest software is configured. Ookla's software needs 'tweaking' (multithreading) to reliably report speeds test that approach or exceeds 100Mb.

We have peering agreements with other ISPs but we can't dictate the packet flow through their own networks or if they choose to statically oversubscribe some infrastructure by design or network dimensioning.

Lastly, the further away the speedtest server you are testing against is from your machine, the more bandwidth-delay product and tcp windoww size plays a part in keeping the variable size of that pipe between you and the speedtest server as full as you can.

TL:DR = lots of factors affect how much bandwidth exist betwwen you and a speedtest server, almost all of them you have no control over, nor does your ISP.

"Customers don’t expect you to be perfect. They do expect you to fix things when they go wrong." Donald Porter – British Airways

The views expressed here are my own and are not reflective of other organisms or organisations.

mercutio

1387 posts

Uber Geek

#773814 3-Mar-2013 09:52

DoomlordVekk: Ian, its the design & the dimensioning of the TCL(now VF) network that says "All things being equal, you should be able to achieve your plan speed across the TCL network".

Its also in the way the speedtest software is configured. Ookla's software needs 'tweaking' (multithreading) to reliably report speeds test that approach or exceeds 100Mb.

We have peering agreements with other ISPs but we can't dictate the packet flow through their own networks or if they choose to statically oversubscribe some infrastructure by design or network dimensioning.

Lastly, the further away the speedtest server you are testing against is from your machine, the more bandwidth-delay product and tcp windoww size plays a part in keeping the variable size of that pipe between you and the speedtest server as full as you can.

TL:DR = lots of factors affect how much bandwidth exist betwwen you and a speedtest server, almost all of them you have no control over, nor does your ISP.

Surely it should be possible to do 100 megabit nationally without multiple threads.

That said there are a few issues with speedtest. The first is that it uses too small files by default, and often doesn't ramp up to big enough files to get an accurate test. Second is that it's not very efficient - even testing on a local gigabit network to another host can be a bit hit and miss with getting full download speed. Seemingly going better with IE than Chrome. Thirdly, there are adequete peering exchanges in NZ that can be used to ensure users get ideal performance. I'm pretty sure all the speed test servers on speedtest.net for New Zealand are peered at APE at least, and most likely the bulk are also peering at WIX. If TelstraClear don't choose to peer and have reduced national performance then to my mind it's a problem in the TelstraClear network rather than with the hosts.

Also to test the "on net" thing you may want to try the Christchurch Snap speed test server as they have some domestic transit via TelstraClear.

Actually one thing that could be effecting this is that there's a Telecom performance issue recently, and most parties are only getting domestic transit to TelstraClear/Telecom via Telecom or TelstraClear, whereas all the other ISP's are peering. Basically the gist is that TelstraClear and Telecom want to charge content providers to reach their customers, and want to charge nationally at above cost price (ie for profit). So circuits are much more likely to be oversubscribed etc. Although afaik, the main oversubcriber atm is Telecom/Chorus, and that's within their own network rather than outside it.

As a test you can try:

telecom (akld): http://speedtest3.telecom.co.nz/

wxc (akld): http://speedtest.wxnz.net/

snap (chch): speedtest.net, go to christchurch

freitasm

BDFL - Memuneh
67451 posts

Uber Geek

Administrator

Trusted

Geekzone

Lifetime subscriber

#773824 3-Mar-2013 10:14

mercutio: Actually one thing that could be effecting this is that there's a Telecom performance issue recently, and most parties are only getting domestic transit to TelstraClear/Telecom via Telecom or TelstraClear, whereas all the other ISP's are peering.

Telecom and TelstraClear stopped peering for free more than five years ago.

These links are referral codes

mercutio

1387 posts

Uber Geek

#773828 3-Mar-2013 10:36

freitasm:
mercutio: Actually one thing that could be effecting this is that there's a Telecom performance issue recently, and most parties are only getting domestic transit to TelstraClear/Telecom via Telecom or TelstraClear, whereas all the other ISP's are peering.

Telecom and TelstraClear stopped peering for free more than five years ago.

Yeah, but more to the point, he was doing a speed test to Orcon. Orcon get domestic transit off Telecom not TelstraClear, so it has to traverse Orcon -> Telecom -> Telstraclear, so he could be being hit by the Telecom issue when using the Orcon server. Although this issue manifests in evenings rather during the day, so if testing during the daytime it should be fine.