dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
2410
Webslinger
Premium Member
join:2004-05-01

1 edit

Webslinger

Premium Member

Teksavvy's speedtest server results often aren't reliable




Hosted by Nexicom



Hosted by Teksavvy



During peak hours and weekends, I often don't trust the results anymore from Teksavvy's speedtest server.

I'm only posting this for people who rely on Teksavvy's speedtest results solely (and yes, I realize the speedboost D3 results are high from the other two tests, but at least those are consistent). During offpeak hours, Teksavvy speedtest server results are similar to the ones I posted from St. Catherines and Nexicom.

sbrook
Mod
join:2001-12-14
Ottawa

sbrook

Mod

I noticed that a long time ago ... I suspect that the number of users is greater than the capacity of the pipe to the machine, or the machine just doesn't have the oomph!
Xuhum
Premium Member
join:2007-10-05
Canada

Xuhum to Webslinger

Premium Member

to Webslinger
20Mbps! not as bad as I had it once, thought my node was actually really congested!, until...

Teksavvy:


Nexicom:


Beanfield Metroconnect:

LINCSAT
Premium Member
join:2003-06-15
Markham, ON

LINCSAT to Webslinger

Premium Member

to Webslinger
Here are my results very reliable









JenSuisUn
Premium Member
join:2006-02-23
Chatham, ON

JenSuisUn

Premium Member

I believe the OP was referring the TekSavvy Speedtest server in Toronto.
Webslinger
Premium Member
join:2004-05-01

Webslinger

Premium Member

said by JenSuisUn:

I believe the OP was referring the TekSavvy Speedtest server in Toronto.

Yes
kabes
join:2010-05-14
Kitchener, ON

kabes to Webslinger

Member

to Webslinger
I've noticed the same. I get faster results from pretty much every other speedtest server in Ontario. For example just now I got 24Mbps on teksavvy and 32Mbps on beanfield. The ping on teksavvy's is usually the lowest, sometimes as low as 5ms on my wired PC.

EssGee
join:2002-01-06
Toronto, ON
Hitron CGN3

EssGee to Webslinger

Member

to Webslinger
Definitely some "variety" in results... (supposed to be 28/1)

TSI


Beanfield




Nexicom

Gone
Premium Member
join:2011-01-24
Fort Erie, ON

Gone to Webslinger

Premium Member

to Webslinger
I noticed this months ago, mentioned something about it in another thread, and then Marc made fun of me and said it was impossible.


TSI Marc
Premium Member
join:2006-06-23
Chatham, ON

TSI Marc

Premium Member

Been trying to chase down what's wrong with it since then.. We did make some changes but I'm with you now.. Something is wrong with it.

It's sitting right in the core though... There's no reason for it.

Gone
Premium Member
join:2011-01-24
Fort Erie, ON

Gone

Premium Member

I will be the first to admit that I couldn't reproduce it on a consistent basis.

tristen1230
join:2010-12-26
Belleville, ON

1 edit

tristen1230 to Webslinger

Member

to Webslinger

Teksavvy
 

Beanfield
 

Nexicom
 
On a 14 Mbps plan. They seem all fairly consistent.

Gone
Premium Member
join:2011-01-24
Fort Erie, ON

Gone

Premium Member

Are you sure you didn't post the same result image three times?

tristen1230
join:2010-12-26
Belleville, ON

tristen1230

Member

Ooops. :P Let me do that again.

Gone
Premium Member
join:2011-01-24
Fort Erie, ON

Gone to Webslinger

Premium Member

to Webslinger
Nexicom:




Teksavvy:




Looks like things are okay right at this very moment, for me anyway.
Webslinger
Premium Member
join:2004-05-01

Webslinger

Premium Member

You need to be testing during prime time (midnight EST is not prime time). Try around 9 p.m. EST on a weekday. And try the St. Catharines' speedtest server for comparison.

Gone
Premium Member
join:2011-01-24
Fort Erie, ON

1 recommendation

Gone

Premium Member

I get better results on Nexicom or Beanfield than I ever have on NRBN.

Midnight typically is primetime, too. Once you hit 2AM things start to drop off. Anyone who's ever suffered node congestion knows this all too well
Webslinger
Premium Member
join:2004-05-01

2 edits

Webslinger

Premium Member

said by Gone:

Midnight typically is primetime, too.

No, I never have problems with the Teksavvy Toronto speedtest server at midnight on a Sunday. And this specific situation has a lot less to do with node congestion than Teksavvy's speedtest server in Toronto being congested.

Try again at 9:15 p.m. EST

If you get bad results from Teksavvy, try the St. Catharines' speedtest server for comparison.

tristen1230
join:2010-12-26
Belleville, ON

tristen1230 to Webslinger

Member

to Webslinger

Teksavvy
 

Beanfield
 

Nexicom
 
Well it is 4:00 on a holiday and here are my results.
Webslinger
Premium Member
join:2004-05-01

Webslinger

Premium Member

Can we please stick to customers in the GTA (it's best to test servers near you) that are subscribed to services faster than 16 Mb/s (to at least stress the server somewhat) down--testing during peak times on Teksavvy's Toronto server?

Mike2009
join:2009-01-13
Ottawa, ON

Mike2009

Member

Why?
Webslinger
Premium Member
join:2004-05-01

1 edit

Webslinger

Premium Member

The premise is that Teksavvy's Toronto speedtest server is choking during peak times. Typically, lower pings (and more reliable results) are based on server distance. And you aren't going to be stressing much of anything with lower download speed limits. If Teksavvy's server taps out at 24 Mbps during peak times, and you have a 16 Mb/s service, whopdeedo if you get between 16 Mb/s and 24 Mb/s from the speedtest results (for example).

sbrook
Mod
join:2001-12-14
Ottawa

1 recommendation

sbrook to Webslinger

Mod

to Webslinger
ALL TekSavvy's Ontario cable customers are handed off through Rogers and Cogeco through TSIs aggregating routers in Toronto at 151 Front and then to routers to the selected route to the destination.

Some routers are "invisible" so we don't see those hops but Nexicom's and TekSavvy's tracert servers both appear to be at 151 Front in Toronto ... the very same place TekSavvy's Ontario cable customers arrive at!

So slightly different routing around the building and hey presto. So the differences between TekSavvy and Nexicom's speedtest results therefore can only be one of a very few things ...

1) the TekSavvy server is slower at queuing up the packets to send down the line either because it's not got enough oomph or other configuration issue

2) or the TekSavvy server is overloaded doing its job

3) because the links / internal routing in the 151 building are slower to TekSavvy's speedtest server than they are to Nexicom's

It doesn't matter where you are in the province ... that's the route.

I'm not sure of the routing from Bell.

(In doing some research I did note that it seems that inbound packets say to a TSI customer in Ottawa leave TekSavvy's aggregating router and then go to Rogers on Bloor Phub! Go figure almst as if they are aggregated now for the upstream!

Gone
Premium Member
join:2011-01-24
Fort Erie, ON

Gone to Webslinger

Premium Member

to Webslinger
said by Webslinger:

If you get bad results from Teksavvy, try the St. Catharines' speedtest server for comparison.

Considering that Start peers with Torix at 151 Front, and the Nexicom and Beanfield servers are also located at Torix, and that NRBN's server (that's the "St. Catharines" one you're talking about) is located on the other side of Lake Ontario (I want to say either 80 King or 15 Church) it makes far more sense for me to use one of the Toronto servers that are on-net with Start than it is for me to use on St. Cats even though I'm in Fort Erie.

This obviously isn't node congestion, I never said it was. Regardless of the result with Teksavvy's own server, midnight is peak time for Internet usage, though it begins to drop off after that time. You can actually see just how much traffic goes on at Midnight here »www.torix.ca/stats.php.

Furthermore, I think we all know what time zone we're all in.
Webslinger
Premium Member
join:2004-05-01

3 edits

Webslinger

Premium Member

said by Gone:

Considering that Start peers with Torix at 151 Front, and the Nexicom and Beanfield servers are also located at Torix, and that NRBN's server (that's the "St. Catharines" one you're talking about) is located on the other side of Lake Ontario (I want to say either 80 King or 15 Church)


Yeah, I know.
quote:
it makes far more sense for me to use one of the Toronto servers that are on-net with Start
I suggest the Toronto speedtest servers, in general, are being hit harder than the one in St. Catharines during the times I'm testing (and at midnight on Sundays I get great results from Teksavvy's speedtest server, so midnight speedtest results don't impress me unless a Teksavvy rep is going step in and tell us that at midnight on Sunday Teksavvy's speedtest server was being hit harder than it was at 9 p.m.).
quote:
This obviously isn't node congestion, I never said it was.
You said that midnight constitutes "primetime" and that "anyone who's ever suffered node congestion knows this all too well".

But I'm not talking about peak times with respect to general internet usage in this thread. I'm referring specifically to times when I suspect speedtest servers (and possibly, routing to them, to a lesser extent) in Toronto are being maxed out.
quote:
Regardless of the result with Teksavvy's own server
No, that's the whole point. This has everything to do with Teksavvy's server (sbrook's points #1 and #2, I suspect).
quote:
midnight is peak time for Internet usage
I suggest midnight is not peak time to be running speedtests on Teksavvy's Toronto server. I could be wrong, but if I am, why are my results (and that of other Teksavvy users I know) typically great at midnight?
quote:
Furthermore, I think we all know what time zone we're all in.
Well, that's reassuring, I guess.
Webslinger

2 edits

Webslinger to sbrook

Premium Member

to sbrook
said by sbrook :
So the differences between TekSavvy and Nexicom's speedtest results therefore can only be one of a very few things ...
Well, except sometimes I get bad results from Nexicom as well at the same time as I do from Teksavvy.
quote:
It doesn't matter where you are in the province ... that's the route.
Yes, but there are more hops and routers along the path, the further away one is situated.

Gone
Premium Member
join:2011-01-24
Fort Erie, ON

Gone to Webslinger

Premium Member

to Webslinger
said by Webslinger:

The Toronto speedtest servers, in general, are being hit harder than the one in St. Catharine's during the times I'm testing (and at midnight on Sundays I get great results from Teksavvy's speedtest server, so midnight speedtest results don't impress me unless a Teksavvy rep is going step in and tell us that at midnight on Sunday Teksavvy's speedtest server was being hit harder than it was at 9 p.m.).

Well, aside from the fact that neither Beanfield nor Nexicom have demonstrated any sort of slowdowns due to capacity, if you use the NRBN server you're throwing a whole wrench into the equation because now you're also limited by the speed of the links (and routers, since it's going to go through Hamilton) to get to NRBN, rather than to Nexicom's or Beanfield's edge router at 151 Front.
said by Webslinger:

I'm not talking about peak times with respect to general internet usage in this thread. I'm referring specifically to times when I suspect speedtest servers (and possibly, routing to them, I guess) in Toronto are being maxed out.

... well, you sort of did when you said Midnight wasn't peak times. It still is, it's just right at the edge. It then drops off until it peaks again at noon. The heaviest traffic is between 8PM and 10:30PM, but it's still comparatively heavy at even midnight.
said by Webslinger:

I suggest midnight is not peak time to be running speedtests on Teksavvy's Toronto server. I could be wrong, but if I am, why are my results typically great at midnight?

Well, it could be any number of things. Either way, Marc and his people are smart. They'll figure it out.
said by Webslinger:

Well, that's reassuring, I guess.

What's less reassuring is that someone doesn't know that Fort Erie, ON is in Eastern Time.
Gone

Gone to Webslinger

Premium Member

to Webslinger
said by Webslinger:

Well, except sometimes I get bad results from Nexicom as well at the same time as I do from Teksavvy.

If that's the case, you might have more serious issues on your hand than just a congested Teksavvy server.
Webslinger
Premium Member
join:2004-05-01

1 edit

Webslinger

Premium Member

said by Gone:

said by Webslinger:

Well, except sometimes I get bad results from Nexicom as well at the same time as I do from Teksavvy.

If that's the case, you might have more serious issues on your hand than just a congested Teksavvy server.

And at the same time get fantastic results from St. Catharines', despite being further away
Webslinger

2 edits

Webslinger to Gone

Premium Member

to Gone
said by Gone :
Well, aside from the fact that neither Beanfield nor Nexicom have demonstrated any sort of slowdowns due to capacity
Alright, what would explain faster results at 9:30 p.m. EST from GTA users using other speedtest servers than from ones located in Toronto?
quote:
well, you sort of did when you said Midnight wasn't peak times
Are you telling me speedtest servers in Toronto are dealing with more traffic at midnight EST than they are at 9 p.m. EST?
quote:
What's less reassuring is that someone doesn't know that Fort Erie, ON is in Eastern Time.

Who might that be?