Speedtest/Spark Throughput

128 views
Skip to first unread message

Matt Strickland

unread,
Jul 18, 2019, 1:46:40 AM7/18/19
to Techies for schools
Hi all,

I assume most schools are on a 1gb up/down connection with N4L?, at least all the ones above 500 students?

Is this a realistic indication of bandwidth capacity? (even peak?)

I've done some testing over the holidays, plugging into port5 on the Fortigate, having profiles removed and running as clean as possible, 11pm tests etc.
I pretty much never hit >500Mbit using speedtest / nperf (350 average, 450peak) - Chorus finds no issues.

What's your speed test result?

Matt

Matt Strickland

unread,
Jul 20, 2019, 6:29:07 AM7/20/19
to Techies for schools
Here is my typical test, either plugged into port5 directly, against 10Gb capacity links, tried Spark Wellington/Auckland, different devices.

Its only download I would expect to see ~750-900 Mbs. Is it just our school?

n4l-test-techies.PNG

Matt

Jake Wills

unread,
Jul 20, 2019, 8:42:22 PM7/20/19
to Techies for schools
Hi Matt,

I just did two speed tests... these are just done on one of our servers that is plugged into core switch which is plugged into fortigate... so not as robust a test as yours...

One with Speedtest by Ookla which is what we normally use...

8433526713.png


And one with nperf which is what it looked like you were using that gave significantly lower results:


3202365284261736-lbhdtg8I.png


Both were done on the same machine within 2 minutes of each other... maybe nperf just reports lower results?
What do you get with https://www.speedtest.net/

Jake

Craig Knights

unread,
Jul 20, 2019, 9:50:35 PM7/20/19
to techies-f...@googlegroups.com
We have had the 1Gb upgrade but see the same as you. It's been slower since Fortigate..  N4L are investigating.  

Far more annoying  is our 4 to 7 Mb upload to Google anything since Fortigate.

1GB takes 40+ min. Etc.  Home on 1Gb is in the order of 30sec down 1min up. That sort of thing.

I think it also slows even our smtp bulk sends as we use GMail for that.

Did I see a Google staff member on here sometimes?

We've just got some VR gear. So doing anything to Youtube with that is going to be painful.

Cheers
Craig.


--
You received this message because you are subscribed to the Google Groups "Techies for schools" group.
To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-sch...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/techies-for-schools/c934880b-7190-4ffa-b8f6-04a335d3493f%40googlegroups.com.

Clayton Hubbard

unread,
Jul 20, 2019, 10:14:22 PM7/20/19
to techies-f...@googlegroups.com
Hi All

Just wanted to acknowledge the comments made so far and we will investigate further into this. 

Clayton

--
Clayton Hubbard
Senior Solutions Architect
0220430155

Peter Lambrechtsen

unread,
Jul 21, 2019, 5:51:11 AM7/21/19
to techies-f...@googlegroups.com
It depends on the path you take and if any of the fortigate hardware is rated above 1gb or not.
Rather than using speedtest I prefer to test a large file download from NZ vs AU vs West coast USA vs UK. Time how long that takes to get an idea of true speed.

I've spent a *LONG* time at Spark when I was there working on speedtests and comparing different hardware against different shapers.

If you are doing uploads you should be shaping the traffic to 500mbit or so since if you try and send at 1gb then you will get packet loss and then tcp resync and it will slow it all down.

All the Spark speed test servers should have been upgraded as it was just about to happen when I left almost a year ago they were busy upgrading them from 1gb to 10gb.

8433526713.png
3202365284261736-lbhdtg8I.png

Matt Strickland

unread,
Jul 22, 2019, 9:41:18 AM7/22/19
to Techies for schools
During the download I also captured some packets - of the small sample of 7880 packets, 7874 were forwarded, 6 were dropped.

Ill need to do more testing and analysis to get a true reading, I'm not really experienced enough to comment, I feel I'm not getting significant packet loss at least during the download phase.

Ill look at downloading a large hosted file and see where that leads.

I also wonder if MTU of 1420 is suitable using the Fortigate. It seemed to suit the Cisco router but could be losing performance?

Matt
To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-schools+unsub...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Techies for schools" group.
To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-schools+unsub...@googlegroups.com.

Peter Lambrechtsen

unread,
Jul 22, 2019, 9:02:48 PM7/22/19
to techies-f...@googlegroups.com
Also you are testing directly to the speedtest server and the traffic isn't getting routed via the proxy server.

If you have a machine with a 10g interface can you directly connect that to the internet bypassing everything.

I assume it's a standard routed IPoE connection rather than PPPoE as a 10gb connection should be delivered over Sparks Carrier Ethernet service.

Plus if you have a Windows host make sure you test with the Windows 10 Speedtest app rather than trying to use a browser. As the browser and python speedtest-cli will always under represent the speeds especially above 500mbit.

In almost all cases I worked on when 1GB Retail UFB launched the problem was always customer routers not up for the job, or desktop computers also not capable. These were cases escalated via 1st/2nd/3rd level support after extensive testing had been done and customers were sure their environment was fine. 
There were only 4 in memory out of hundreds where the LFC had misprovisioned the service as in the early days it was manually done by the smaller LFCs. One was also where a GPON SFP had partially failed and they were getting 80% packet loss at the OLT. 11 other customers on a number of other RSPs were also impacted but it had been faulty for 7+ months and no other cases had been logged about the issue.

The latest versions of iperf3 with a well connected test endpoint is your friend. Plus tcp and udp testing to see if it's window or packet fragmentation related.

While they were fun times. I don't miss arguing and spending far too long with sometimes highly technical people convinced they "weren't getting the service they were paying for" to only say "sorry it was actually this piece of equipment in my environment"


To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-sch...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Techies for schools" group.
To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-sch...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Techies for schools" group.
To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-sch...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/techies-for-schools/50cf3c62-f798-4654-9162-e86f341ba56a%40googlegroups.com.

Steve Voisey

unread,
Jul 29, 2019, 9:35:53 PM7/29/19
to techies-f...@googlegroups.com
I think it would be good if N4L could advise on a couple of things:
- which speed tests to use and why? I get very different results using fast.com (the netflix one) and speedtest.net
- also, what speeds should be expected. 
- finally, what sort of splash page should I expect? Nine time out of ten I get something like this:
image.png



--
Regards,
Steve Voisey

Glenn Magill

unread,
Jul 30, 2019, 9:30:59 PM7/30/19
to Techies for schools
Good morning Steve, 

N4L have a guide on how to run a speed test which can be found here
When you go to speedtest.net I recommend you select 'change server' and select a Spark server that is closest to you. Selecting a Spark hosted server will likely provide a more accurate result for raw bandwidth, as traffic will not be traversing long distances or other ISP's to reach the test server.

My team generally test speed using iPerf rather than web hosted speed test servers, as we believe the results are more reliable for throughput. The iPerf server we use is in our data center, and is a command line tool.
I'm not aware of any public iPerf test servers available on Spark sorry Steve.

In terms of Speed we use 80% of circuit speed as a guideline, if throughput falls below this, we may investigate with Spark and the LFC (local fiber company).

Regarding the splash page you are getting, give us a call on 0800 LEARNING and we will take a look with you to see what is happening there.

Regard
Glenn Magill
Engineering Team Lead
The Network for Learning Ltd
0800 LEARNING




To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-schools+unsub...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Techies for schools" group.
To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-schools+unsub...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Techies for schools" group.
To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-schools+unsub...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Techies for schools" group.
To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-schools+unsub...@googlegroups.com.


--
Regards,
Steve Voisey

Donovan Marshall

unread,
Aug 18, 2019, 12:50:13 PM8/18/19
to techies-f...@googlegroups.com
Jumping in a bit late here but as we seem to have similar issues...

We noticed a slowdown of Google Drive uploads since our Fortigate went in, I'm still seeing reduced speeds -sub 20Mbps, and not consistent it jumps up and down- with uploads to Google Drive at 12:44am with no one else on the network. This is on wired desktops so not a Wi-Fi issue locally.

It's made working on Chromebooks interesting this year as there seems to be a lot of lag with these on the school network that doesn't happen when I'm working at home with my 20Mbs down / 6Mbs up connection at home.

Donovan


Pete Mundy

unread,
Aug 18, 2019, 8:04:57 PM8/18/19
to techies-for-schools@googlegroups.com schools
Hiya Donovan

FWIW, I find that Google Team Drive uploads perform much faster when QUIC traffic is allowed through. I haven't paid that much attention to normal Google Drive uploads, but I'd be willing to bet the data goes to the same servers.

If you're blocking QUIC it might be worth asking for it to be unblocked so that you can do a comparative test. If you find it helps then let us know :)

Pete

> On 19/08/2019, at 12:49 AM, Donovan Marshall <dmar...@kaitaiacollege.school.nz> wrote:
>
> Jumping in a bit late here but as we seem to have similar issues...
>
> We noticed a slowdown of Google Drive uploads since our Fortigate went in, I'm still seeing reduced speeds -sub 20Mbps, and not consistent it jumps up and down- with uploads to Google Drive at 12:44am with no one else on the network. This is on wired desktops so not a Wi-Fi issue locally.
>
> It's made working on Chromebooks interesting this year as there seems to be a lot of lag with these on the school network that doesn't happen when I'm working at home with my 20Mbs down / 6Mbs up connection at home.
>
> Donovan
>
>
> On Sun, 21 Jul 2019 at 09:50, Craig Knights <craig....@gmail.com> wrote:
> We have had the 1Gb upgrade but see the same as you. It's been slower since Fortigate.. N4L are investigating.
>
> Far more annoying is our 4 to 7 Mb upload to Google anything since Fortigate.
>
> 1GB takes 40+ min. Etc. Home on 1Gb is in the order of 30sec down 1min up. That sort of thing.
>
> I think it also slows even our smtp bulk sends as we use GMail for that.
>
> Did I see a Google staff member on here sometimes?
>
> We've just got some VR gear. So doing anything to Youtube with that is going to be painful.
>
> Cheers
> Craig.
>
>
> On Sun, 21 Jul 2019, 8:42 AM Jake Wills, <jake....@kc.school.nz> wrote:
> Hi Matt,
>
> I just did two speed tests... these are just done on one of our servers that is plugged into core switch which is plugged into fortigate... so not as robust a test as yours...
>
> One with Speedtest by Ookla which is what we normally use...
>
> <8433526713.png>
>
>
>
> And one with nperf which is what it looked like you were using that gave significantly lower results:
>
>
>
> <3202365284261736-lbhdtg8I.png>
>
>
> Both were done on the same machine within 2 minutes of each other... maybe nperf just reports lower results?
> What do you get with https://www.speedtest.net/
>
> Jake
>
> On Saturday, 20 July 2019 18:29:07 UTC+12, Matt Strickland wrote:
> Here is my typical test, either plugged into port5 directly, against 10Gb capacity links, tried Spark Wellington/Auckland, different devices.
>
> Its only download I would expect to see ~750-900 Mbs. Is it just our school?
>
>
>
> Matt
>
>
>
>
> On Thursday, 18 July 2019 13:46:40 UTC+12, Matt Strickland wrote:
> Hi all,
>
> I assume most schools are on a 1gb up/down connection with N4L?, at least all the ones above 500 students?
>
> Is this a realistic indication of bandwidth capacity? (even peak?)
>
> I've done some testing over the holidays, plugging into port5 on the Fortigate, having profiles removed and running as clean as possible, 11pm tests etc.
> I pretty much never hit >500Mbit using speedtest / nperf (350 average, 450peak) - Chorus finds no issues.
>
> What's your speed test result?
>
> Matt
>
> --
> You received this message because you are subscribed to the Google Groups "Techies for schools" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-sch...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/techies-for-schools/c934880b-7190-4ffa-b8f6-04a335d3493f%40googlegroups.com.
>
> --
> You received this message because you are subscribed to the Google Groups "Techies for schools" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-sch...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/techies-for-schools/CAMueQa0v1gnm5kOKoASrdc3z%3DAvyx5QZwUzMWrC_60KAQ4t%2BJw%40mail.gmail.com.
>
>
> --
>
>
> Donovan Marshall
> ICT Technician
>
> Kaitaia College
> www.kaitaiacollege.school.nz
> dmar...@kaitaiacollege.school.nz
>
>
> --
> You received this message because you are subscribed to the Google Groups "Techies for schools" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to techies-for-sch...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/techies-for-schools/CANBFP8ZZuj%3DhwWqHi0ipZb8wgAuMFon8MYih%2BJVAPQy%3DtXfXzg%40mail.gmail.com.

signature.asc

Craig Knights

unread,
Aug 18, 2019, 8:36:34 PM8/18/19
to techies-f...@googlegroups.com
I've had a call with N4L and Spark on the line..  they could not get it to speed up.  

It's very weird. We still get 4Mb up.

Peter Lambrechtsen

unread,
Aug 18, 2019, 8:49:12 PM8/18/19
to techies-f...@googlegroups.com
I assume the traffic isn't being decrypted as part of Layer 7 inspection by the Fortigates, and that you're not being routed via the Cisco CWS Proxy server?

On a direct internet connection with only NAT you should at least be getting 20-30mbit depending where in the country you are.

The further south you go the slower it will be due to tcp round trip time. But 4mbit seems very unusual.

Just shooting the breeze here but what happens if a secondary service with just standard Spark Retail Broadband is provisioned to the second port on the ONT?

Josh Angel

unread,
Aug 18, 2019, 9:10:29 PM8/18/19
to techies-f...@googlegroups.com

I’ve seen it hit 120+ Mbps, but average around 60-80Mbps on a quieter day, also seen right down to 15-20Mbps on others. This is all observed with my rClone server (great project if you’ve not seen it before).

 

Running a job right now, it’s showing ~70Mbps upload.

 

 

It could be that daily upload limits are being reached? Some days I’ve swore I’ve seen our uploads throttle…but never fully cut.

 

https://support.google.com/a/answer/7338880

 

----

 

Josh Angel   |   Systems & Network Manager

JAn...@mags.school.nz

help...@mags.school.nz

Alberton Avenue, Mount Albert, Auckland 1025

Pete Mundy

unread,
Aug 18, 2019, 9:14:52 PM8/18/19
to techies-f...@googlegroups.com
Dear List Admins,

Could a list administrator please reach out to Elim's support desk?

Someone has subscribed their ticketing address to the list and their support team is unsure how to unsubscribe the address.

I got an automated response to my posting earlier this morning; see below.

sup...@elimchristiancollege.zendesk.com

Contact is Steven. I can provide personalised email if you contact me off-list.

Thanks,
Pete


> Begin forwarded message:
>
> From: Elim Christian College <sup...@elimchristiancollege.zendesk.com>
> Subject: [Request received] [techies-for-schools] Re: Speedtest/Spark Throughput
> Date: 19 August 2019 at 8:30:04 AM NZST
> To: Pete Mundy <pe...@fiberphone.co.nz>
> Reply-To: Elim Christian College <support...@elimchristiancollege.zendesk.com>
>
> ##- Please type your reply above this line -##
> Your request (#8955) has been received and is being reviewed by our support staff. Please allow 24 hours for a reply as we are very busy at the moment.
>
> To add additional comments, reply to this email or click the link below:
> http://elimchristiancollege.zendesk.com/hc/requests/8955
>
>
>
> Pete Mundy
> Aug 19, 8:30 AM NZST
signature.asc
Reply all
Reply to author
Forward
0 new messages