I work for a ISP and we generally use iperf (have our own server) to do
bandwidh tests. However this is done only of business customers.
Residential customers get best effort.
On 10/24/15 11:09, STR . wrote:
The other benefit is that my ISP, amongst a bunch of local ISPs here, and I'm
sure in other countries too, use speedtest.net to corroborate their bandwidth
offering.
Aggregate results against speedtest.net servers makes for arguably stronger
evidence than my own nuttcp instances even if collocated at the ISP DC.
-----Original Message-----
From: Adrian Popa [mailto:[email protected]]
Sent: Friday, October 23, 2015 12:56 PM
To: Erik Taraldsen <[email protected]>; STR . <[email protected]>
Cc: [email protected]
Subject: Re: [smokeping-users] [BULK] RE: New probe speedtest
The only benefit speedtest has over other testing solutions (iperf, ftp) is the
existence of a ton of test servers. This means you can measure (with best
effort) available bandwidth between two points on the internet (not your own
network).
@STR: you can set up multiple probes and assign each one a different offset so
that you'll be sure the speedtest don't overlap and you don't get slowed down
by your uplink. Or specify the offset as part of the host configuration (should
work)
On 10/23/15 09:21, Erik Taraldsen wrote:
Hi!
If you have the ability to set up an dedicated test server I have
implemented an speed test for smokeping using nuttcp. Given that
nuttcp has more granular controll than speedtest.net you can do things
as create traffic streams for different traffic classes. For example
using udp to model SIP RTP, and throughput in high vs low priority
que.
The code is written for my own test purposes using my fairly limited
skill set so the code won't impress anybody. But if you have the
need, just contact me and I'll happily share.
-Erik
On Fri, Oct 23, 2015 at 2:40 AM, STR . <[email protected]> wrote:
Adrian,
It’s probably not enough of a sample size but if I just use 1
speedtest host local to my city it shows the correct download speeds as I’d
suspected.
This is evidenced by
http://wrt.asuscomm.com:81/smokeping/?target=Speedtest.Download.Tata_
Technologies_Pune_Down
Looking at the Last 30 hours or Last 10 day graphs you can see where
I configured it to use just 1 host, then multihost, then back down to 1 host.
It could well be a speedtest.net issue with me hitting multiple
servers and they just not liking it.
Obviously the probe is reporting the correct values given to it by
speedtest-cli
Also any idea why the Y axis pulls “var units” from Smokeping.pm
unlike other multihost graphs?
Could these 2 issues be related?
- S
From: smokeping-users
[mailto:[email protected]]
On Behalf Of STR .
Sent: Thursday, October 22, 2015 12:02 AM
To: 'Adrian Popa' <[email protected]>;
[email protected]
Subject: Re: [smokeping-users] [BULK] RE: New probe speedtest
“var units” is being pulled from line 892 in smokeping.pm
# if there are multiple units ... lets say so ...
if ($ProbeUnit ne $probe->ProbeUnit()){
$ProbeUnit = 'var units';
}
Any idea why only Speedtest multihost graphs pull that value?
- S
From: Adrian Popa [mailto:[email protected]]
Sent: Tuesday, October 20, 2015 12:35 PM
To: STR . <[email protected]>; [email protected]
Subject: Re: [BULK] RE: [smokeping-users] New probe speedtest
I don't know why the measurement unit is not correct for multigraphs,
but I have tested the plugin and it can correctly display bandwidths
of hundreds of Mbps. Note, that speedtest-cli values are generally
lower than what you get from the speedtest web interface. Also, they will
likely vary in time.
On 10/19/15 21:29, STR . wrote:
Thanks for a great probe!
I think there may be some minor issues still.
I don’t see any of the hosts show me more than 2Mb/s download, which
is incorrect as it should be close to 7.5 – 8 Mb/s
I understand the upload tests might be skewed given the low upload ADSL has.
Some more testing with just the speedtest node closest to me should
shed some light on this.
On the speedtest multihost graphs, only on the main overview page,
the legend for Y axis displays “var units” instead of “bps”.
Example:
https://dl.dropboxusercontent.com/u/314525/MultiHost_mini.png
When I open the multihost page, the Y axis is correctly called “bps”.
Example:
https://dl.dropboxusercontent.com/u/314525/MultiHost_last_10800.png
Why would multihost only change it for speedtest, none of the other
multihost graphs display this issue?
Example:
https://dl.dropboxusercontent.com/u/314525/MultiHost_DNSmini.png
- S
From: Adrian Popa [mailto:[email protected]]
Sent: Monday, October 19, 2015 1:23 PM
To: STR . <[email protected]>; [email protected]
Subject: Re: [BULK] RE: [smokeping-users] New probe speedtest
Sweet, Thanks for your contribution!
On 10/17/15 03:57, STR . wrote:
Of course, it’s been so long I touched Smokeping configurations I
forgot multihost and I use it for DNS!
So get and choose servers in your city/country list from speedtest-cli:
speedtest-cli --list|grep -w Germany
--
TELEKOM ROMANIA
Adrian Popa
Expert Engineer
Applications, Security & EMS
35-37 Calea Victoriei, Palatul Telefoanelor
Mobile: +40 766 673 290
Phone: +40 21 400 7567
_______________________________________________
smokeping-users mailing list
[email protected]
https://lists.oetiker.ch/cgi-bin/listinfo/smokeping-users
.
--
TELEKOM ROMANIA
Adrian Popa
Expert Engineer
Applications, Security & EMS
35-37 Calea Victoriei, Palatul Telefoanelor
Mobile: +40 766 673 290
Phone: +40 21 400 7567
--
TELEKOM ROMANIA
Adrian Popa
Expert Engineer
Applications, Security & EMS
35-37 Calea Victoriei, Palatul Telefoanelor
Mobile: +40 766 673 290
Phone: +40 21 400 7567
_______________________________________________
smokeping-users mailing list
[email protected]
https://lists.oetiker.ch/cgi-bin/listinfo/smokeping-users
