Koozali.org: home of the SME Server

Load Testing BT Infinity UK

Offline gbentley

  • ****
  • 482
  • +0/-0
  • Forum Lurker
    • Earth
Load Testing BT Infinity UK
« on: October 19, 2017, 02:00:01 PM »
A client has roof top radio for internet access which is really quick [around 80Mb down / 55 up] but costly. Now that there is BT Infinity [FTTC] on the Exchange [its taken years] they have gone ahead and got an IP phone system installed, which uses the fax line + Infinity.

There are a few lines left and we could add another Infinity for data however we are shifting about 160GB / month and from what I have seen at other sites, Infinity doesn't always deliver in terms of speed [no idea about bandwidth]

I need a way to make an assessment on the existing line, speed tests and some kind of averaged out transfer rates / throughput etc

I can setup a box to up/down load data but I am no shell script guru. I have used this ;

https://linhost.info/2013/10/download-test-files/

to try and make an assessment but would want to also test uploading and gather all the data over say a 48 hr period?

Right now I have no idea how to achieve this. If anyone can ping in with suggestions or even comment on their experience with Infinity [FTTC] in the UK it would be really appreciated :)
 
"If you don't know what you want, you end up with a lot you don't."

Offline ReetP

  • *
  • 3,732
  • +5/-0
Re: Load Testing BT Infinity UK
« Reply #1 on: October 19, 2017, 06:37:38 PM »
Plenty of open source speed test tools about. Vnstat springs to mind but there was another which the old grey cells have forgotten

If FTTC doesn't cut it the more expensive option is EFM - Ethernet First Mile

No contention (which is usually the issue) and AFAIAA symetrical speeds.

Claranet were very responsive to my queries. BT and Zen (surprisingly) were not!
...
1. Read the Manual
2. Read the Wiki
3. Don't ask for support on Unsupported versions of software
4. I have a job, wife, and kids and do this in my spare time. If you want something fixed, please help.

Bugs are easier than you think: http://wiki.contribs.org/Bugzilla_Help

If you love SME and don't want to lose it, join in: http://wiki.contribs.org/Koozali_Foundation

Offline gbentley

  • ****
  • 482
  • +0/-0
  • Forum Lurker
    • Earth
Re: Load Testing BT Infinity UK
« Reply #2 on: October 30, 2017, 10:50:23 AM »
These are some of the tests I have tried from then desktop;
speedof.me
testmy.net
speedtest.net
bandwidthplace.com
speedtest.btwholesale.com
magnet.speedtestcustom.com

and from cli

Location: Dallas
wget -O /dev/null http://speedtest.dal01.softlayer.com/downloads/test100.zip

Location: Netherlands
wget -O /dev/null http://mirror.nl.leaseweb.net/speedtest/1000mb.bin

Location: Tokyo
wget -O /dev/null http://speedtest.tokyo.linode.com/100MB-tokyo.bin

Location: London
wget -O /dev/null http://speedtest.london.linode.com/100MB-london.bin

Location: France
wget -O /dev/null http://proof.ovh.net/files/100Mb.dat

What I am aware of though is that speed tests alone don't tell you very much. I am certainly no network guru but see words like 'capacity' and 'throughput' banded about in discussions.

AFAIUI 'Throughput' is just the amount of data you can down/upload in a given time and this is usually given in Mbps and that's exactly what is reported by the likes of magnet.speedtestcustom.com - but this only refers to data collected at that specific time, from that specific location. It makes me wonder if this is useful at all because most small networks with say 10-20 staff may be downloading and uploading data from/to different locations simultaneously and at different times of the day.

I suspect I would need to setup multiple up/downloads simultaneously and possibly of mixed traffic types [http + smtp + imap] and to sustain this for say 24 hours and then be able to make sense out of the collected data?

Hhmm, this is turning out to be harder than I first thought....

« Last Edit: October 30, 2017, 11:12:32 AM by gbentley »
"If you don't know what you want, you end up with a lot you don't."

Offline ReetP

  • *
  • 3,732
  • +5/-0
Re: Load Testing BT Infinity UK
« Reply #3 on: October 30, 2017, 01:42:44 PM »
What I am aware of though is that speed tests alone don't tell you very much. I am certainly no network guru but see words like 'capacity' and 'throughput' banded about in discussions.

AFAIUI 'Throughput' is just the amount of data you can down/upload in a given time and this is usually given in Mbps and that's exactly what is reported by the likes of magnet.speedtestcustom.com - but this only refers to data collected at that specific time, from that specific location. It makes me wonder if this is useful at all because most small networks with say 10-20 staff may be downloading and uploading data from/to different locations simultaneously and at different times of the day.

I suspect I would need to setup multiple up/downloads simultaneously and possibly of mixed traffic types [http + smtp + imap] and to sustain this for say 24 hours and then be able to make sense out of the collected data?

Hhmm, this is turning out to be harder than I first thought....

Yup - to get an accurate representation you need to run a continual assessment over a period of time. However, the first thought is that you clearly don't shunt 160Gb in one go - it's over a month. Also running it on your desktop probably isn't that representative either - can you stick some monitoring on your actual firewall or gateway ?

Some very rough maths.... assuming a 5 day week that means around 8Gb per day and an based on 8 hours per day that is a rough average of 1Gb per hour which is 17 megabytes a minute, which is about 136 megabits per minute, giving around 2.25 megabits per second (I think !!!!). If your line is say 80/55 megabits then on average it should be fine ??? You should calculate this a bit more accurately, and measure the actual data flow with something like vnstat to make sure you have the right numbers.

Your issue would be with burst speeds - someone wanting a 5Gb iso file NOW. That will be affected by contention ratio on the line - remember you share that 80Mb with a load of other people as well. At 3 am you'll get 80mb but you won't get it continuously in the middle of the day when loads of people are online. So do check the contention ratio on the package offered. Your ISP may also throttle or prioritise certain types of traffic - probably not easy to check that one, and they may not like to admit whether they do :-)

A quick google revealed this page (I am sure there are many more) with some interesting tools:

https://www.serverdensity.com/monitor/linux/how-to/

http://www.binarytides.com/linux-commands-monitor-network/

The other one I was thinking where you can run a client and server and actually push/pull files/data is called netperf - https://github.com/HewlettPackard/netperf

You could probably script/cron it so send regular amounts of data over a period and see what happens. However, remember that you will also be fighting with whatever other data is already flowing.... so it is hard to get a real test unless you can test the line with no other traffic.

HTH (I'm no networking guru!!!!)

B. Rgds
John
...
1. Read the Manual
2. Read the Wiki
3. Don't ask for support on Unsupported versions of software
4. I have a job, wife, and kids and do this in my spare time. If you want something fixed, please help.

Bugs are easier than you think: http://wiki.contribs.org/Bugzilla_Help

If you love SME and don't want to lose it, join in: http://wiki.contribs.org/Koozali_Foundation

Offline gbentley

  • ****
  • 482
  • +0/-0
  • Forum Lurker
    • Earth
Re: Load Testing BT Infinity UK
« Reply #4 on: November 03, 2017, 07:00:09 PM »
Thanks for replies :)

Could you comment on whether the below is valid / meaningful?

As a preliminary test I used this tool ;

https://download.peplink.com/files/B...est%20v3.1.exe

on a Windows 7 Pro desktop PC with a 1GB nic. I opened up multiple instances of the tool and selected the 85Mb test file via Google in each case apart from Test4 where I used the following 100Mb test file locations;

http://speedtest.tokyo.linode.com/100MB-tokyo.bin
http://speedtest.london.linode.com/100MB-london.bin
http://speedtest.newark.linode.com/100MB-newark.bin
http://speedtest.atlanta.linode.com/100MB-atlanta.bin
http://speedtest.dallas.linode.com/100MB-dallas.bin
http://speedtest.fremont.linode.com/100MB-fremont.bin

I added the individual results together to get the total throughput as well as the finish times of the first and last process.

Infinity Test 1: 85Mb x 3
1st Finished: 25s
Last Finished: 25s
Throughput: 66.89Mbps

Infinity Test 2: 85Mb x 6
1st Finished: 49s
Last Finished: 51s
Throughput: 67.35Mbps

Infinity Test 3: 85Mb x 9
1st Finished: 69s
Last Finished: 71s
Throughput: 71.74Mbps

Infinity Test 4: 100Mb x 6
1st Finished: 37s
Last Finished: 77s
Throughput: 77.77Mbps

WiFi Test 1: 85Mb x 9
1st Finished: 76s
Last Finished: 78s
Throughput: 60.72Mbps

WiFi Test 2: 85Mb x 6
1st Finished: 56s
Last Finished: 52s
Throughput: 59.58Mbps

WiFi Test 3: 85Mb x 3
1st Finished: 27s
Last Finished: 28s
Throughput: 59.7Mbps

WiFi Test 4: 100Mb x 6
1st Finished: 39s
Last Finished: 107s
Throughput: 64.83Mbps

Screen Grabs
https://jpegbay.com/gallery/006581264-1.html#1

To clarify the Infinity account has only VOIP phones on and there was low traffic during testing whilst on the WiFi account the entire office activity was in full swing. That said the results aren't nearly as different as I expected.

Next I will be testing multiple uploads when I can come up with a decent test method.

One last thing; would running multiple tests on a single PC be affected by the limitations of the single 1GB nic?
"If you don't know what you want, you end up with a lot you don't."

Offline ReetP

  • *
  • 3,732
  • +5/-0
Re: Load Testing BT Infinity UK
« Reply #5 on: November 03, 2017, 07:18:26 PM »
Sounds like you are getting around 65 megabits/sec (divide by 8 to get your bytes) on average. Only you can tell if that is enough :-)

Doing multiple transfers isn't likely to prove much... they'll just add up to around 65mb....

Using one bigger test file is better. The test files have random data so can't be compressed and therefore give a better idea of speed (takes me back to my 2400 baud modem days!).

It also reduces the chance of caching affecting things. Try a few Gb files as well....

From a quick Google, Gigabit theoretically does 1024 megabits/sec which with overhead gives a usable 118 megabytes/sec. The limitations are going to be your broadband (at about 8 megabytes/sec), not your Gigabit local ethernet. And probably your disk read/write speeds too :-)

HTH.

John
...
1. Read the Manual
2. Read the Wiki
3. Don't ask for support on Unsupported versions of software
4. I have a job, wife, and kids and do this in my spare time. If you want something fixed, please help.

Bugs are easier than you think: http://wiki.contribs.org/Bugzilla_Help

If you love SME and don't want to lose it, join in: http://wiki.contribs.org/Koozali_Foundation