Kitz Forum
Broadband Related => FTTP Rollout => Topic started by: Bestgear on October 10, 2021, 06:56:17 PM
-
Hi Guys
Just wondered what is considered as a sensible way to measure FTTP throughput given that most of the usual suspects hosts are unlikely to have sufficient bandwidth to run your line flat out.
I find that Vodafone (two that are offered by Speedtest.net (app) are up there, but are variable.
All my cable is cat6 (checked with a certifier), and I am using Untangle in a VM (intel nics) and get a respectable performance - see attached image.
I have also tested with a PC directly into the ONT, so no router, and it is not any better for me.
I have not unboxed the BT provided router so cant use their own "account" speed test.
Thoughts?
-
I think the only practical way is to do it between 3am and 6am when less people are likely to be online.
Even then, I've seen faster speeds in real-world usage on Three 5G than any speed test I've done, but that's probably partly down to the difference in a mobile network vs a hardline.
-
You could try speedtest.aa.net.uk, it’s not so heavily used, I would assume, but probably better to pick an odd time, as previous posters said. You really need a speed-tester that is in your own ISP’s network, ideally, so that you won’t have to cross the wider internet of course.
-
What happens if you follow the "if you are testing 500/1000mbps product, click here" ?
-
I tend to use:
- iperf to a server which has either at least a 1Gbit or 10Gbit connection. Oracle Cloud for example is one place I use, as I use their service
- nperf.com
- speedtest.net
- DSLreports or Waveform's Bufferbloat Test (https://www.waveform.com/tools/bufferbloat)
- Various UK mirror servers for ISO images like CentOS (better during early hours of the morning I guess though)
-
What happens if you follow the "if you are testing 500/1000mbps product, click here" ?
Some guidance appears... hehe
-
You could try speedtest.aa.net.uk, it’s not so heavily used, I would assume, but probably better to pick an odd time, as previous posters said. You really need a speed-tester that is in your own ISP’s network, ideally, so that you won’t have to cross the wider internet of course.
Thanks - will give it a try....I find that BT Wholesale tester bad during the day, but evenings and weekends are better for obvious reasons.
I did look at thinkbroadbands tool - but was crap for me... I think photoshop would be needed to get any decent speed recorded ???
-
Great tips here - clearly from experience! Thanks for these.
Has anyone used - and do they rate - the speed tester that is presented within the BT portal?
As I am not using a HomehUb - I cannot use it - and according to the BT support staff they cannot see line speeds without a Homehub either... which I am somewhat sceptical about.
At least with FTTP there is no DLM to worry about messing your speeds up...
D
-
You could try speedtest.aa.net.uk, it’s not so heavily used, I would assume, but probably better to pick an odd time, as previous posters said. You really need a speed-tester that is in your own ISP’s network, ideally, so that you won’t have to cross the wider internet of course.
Need to try it when the kids (across the UK) go to sleep....
-
Perhaps try fast.com (https://fast.com/) ?
-
Need to try it when the kids (across the UK) go to sleep....
Fortunately Fortnite, COD and Roblox doesn't use a lot of bandwidth. ;)
-
Perhaps try fast.com (https://fast.com/) ?
Not great!
-
Fortunately Fortnite, COD and Roblox doesn't use a lot of bandwidth. ;)
Aye, thats about the extent of it!
-
Hi
I am seeing real speed variations in testing - including many well below the pitiful promise of 450Mb.....
Anyone know what the process is in respect of speed issues with BT/OR?
I guess I need to - firstly - put a homehub at the end of the fibre and retest using the myBT portal? Does that provide any direct speed test that excludes the inherent latency and capacity issues of using another public service? I kinds hoped it could test from my homhub to the exchange that would prove the link.
Thoughts?
-
Hi
The only true way to test would be as suggested already, and test against a server you are in control off and which would give you an accurate ping (ping and jitter are calculated best guess as the test servers cannot ping directly to you)
You could download a 1GB non compressible file and time it to calculate
The BT HH speed test is your first picture, here is mine for comparison on hardwired to pc (FTTP 300/50) (FTTC 80/20 previously showed faster but computer appears to be faster on FTTP)
Looks low and untrue but picture 2 is taken from iphone on wireless...
BT always state you will not have a full 900mbps all the time, and it will vary... so what is not working on your network that you need full bandwidth - perhaps look at full leased line
Many thanks
John
-
I would try it with the BT Hub. The My BT app has a 'test my speed' option under 'More'.
I would also be interested in what version firmware, the hub is using, as it may not be the latest and you will need to wait for BT to remotely update it.
-
The only true way to test would be as suggested already, and test against a server you are in control off and which would give you an accurate ping (ping and jitter are calculated best guess as the test servers cannot ping directly to you)
Thanks - its very hard to know really.... I think I need to get the homehub out and try then speak to them next week.
You know fibre should be there or not - no dlm so no variations in speed in theory excluding exchange congestion... and I would hope that would not be an issue with only small numbers of subscribers currently.
I notice too that my IP location is not in my town with FTTP - whereas the IP location on FTTC was a couple of hundred miles away so I can only assume each exchange has a new backbone connection.... which it must have for bandwidth I think....anyone else spotted that?
D
-
You know fibre should be there or not - no dlm so no variations in speed in theory excluding exchange congestion... and I would hope that would not be an issue with only small numbers of subscribers currently.
That's far from reality I'm afraid.
If you want full speed tests at all times of day and full single thread speeds then you're going to be paying a lot more than BT charge for a residential service.
There's PON contention (your splitter node), exchange congestion and backhaul congestion.
That's just within BT's realm before you hit the wider internet.
I've been following 2 recent online complaints with BT FTTP where they have refused to help customers on the 900Mb package who are hitting the guaranteed minimum which is around 450Mb/s.
Both cases hit a deadlock, went to the ombudsman and the ombudsman has ruled in BT's favour.
BT won't entertain your complaint until you connect the Smart Hub 2.
It has a built in speed tester that specifically tests throughput to the Hub. It rules out any and all hardware at the customer end like underpowered PC's/laptops/phones or things like problematic Ethernet adapters.
There's no DLM to affect the sync speed but it's very much a residential service and expectations should match that.
It's a contended service and is advertised/sold/priced at such.
You aren't likely to get far with any complaints about speeds under the minimum guarantee.
It also isn't a pitiful guarantee. What BT guarantee you is considerably higher than what OpenReach guarantee BT.
It's higher than I've seen from any other residential OpenReach FTTP provider.
-
What J0hn said.
To get a ‘guarantee’ of full link speed to your ISP, you need a ‘leased line’ - which is dedicated to you, and costs a fortune per month. My ISP, Andrews and Arnold, for example, will happily sell you one, ultimately provided by BT, will charge you god knows what, and you can have any speed you want. But even that won’t give you what you really want/need, a speed guarantee across the wider internet as the whole internet and servers at the far end is/are not under the control of your ISP. AA will only guarantee ‘not to be the bottleneck’ themselves, with best efforts and within reasonable limits.
But shared PON is no good for you if you want even more speed than say 450 Mbps or whatever. Look at what Carl has achieved for example, speed measurement of what was it 2.1 Gbps downstream ? - some months back.
-
There's the peering and transit of the ISP to consider too, the remote server's capabilities (e.g. 1Gbit port) as well as the network that the remote server is on (their peering, transit, etc). Anything along the route that the specific connection takes. There's many factors which may result in a 'speed test' that doesn't live up to how you thought it would.
For example, my connection with Cerberus to my OVH server, even on 'ultimate bandwidth', has congestion issues somewhere along the route. However, if I go via another server, such as Mullvad (VPN) to reach the OVH server then I have no trouble with some apparent congestion issue.
-
I have found https://nperf.com (http://nperf.com) pretty much maxes out my 900/100 line at most times, when it doesn't I think its contention rather than anything else as there has been quite a large take up since we went live 5 months back.
Here is one run just this minute - https://pasteboard.co/4fDz9kTz5cRv.png
-
I have found https://nperf.com (http://nperf.com) pretty much maxes out my 900/100 line at most times, when it doesn't I think its contention rather than anything else as there has been quite a large take up since we went live 5 months back.
Here is one run just this minute - https://pasteboard.co/4fDz9kTz5cRv.png
Thanks for that - looks like a good tester to stay with.
I am not getting the throughput you have seen, but thats another story. What sped of PC are you using to test and what router are you using?
-
I've been following 2 recent online complaints with BT FTTP where they have refused to help customers on the 900Mb package who are hitting the guaranteed minimum which is around 450Mb/s.
Both cases hit a deadlock, went to the ombudsman and the ombudsman has ruled in BT's favour.
That is interesting.... even with my testing, the speed does vary - with 850ish being seen now and again...even on a pc that the following day will report half that. The BT Wholesale test is variable... upload speed always above the "promise", which has been widely reported.
Been using multiple pcs at the same time (all targeting different test servers and even different test apps/sites) to see what the UTM reports - and it is still capped at pretty much the best one PC can manage (on teh day) so that kinds proves its not internal to the house.
BT won't entertain your complaint until you connect the Smart Hub 2.
It has a built in speed tester that specifically tests throughput to the Hub. It rules out any and all hardware at the customer end like underpowered PC's/laptops/phones or things like problematic Ethernet adapters.
Yes- thats a job for today - had a brief play but it would not allow me to set the LAN subnet to be a 10.0.0.0/16 which I need to slot the thing in, so may need to live with 192.x.x.x for a week of testing which is a bugger for all other devices I have. So - question - if you use the "built in speed tester) where does that run "to"? - the local exchange? Cant see anything else giving much different speeds.
Anyone else run a UTM in a HyperV VM? I have used Untangle for years and is totally stable in a VM. Intel i350 nics, running off a passthrough SSD. Changing cpu allocation (default 2 vcpus, been to 8) and memory (usually 6gb but been to 12) as well as changing VMQ, IPsec offloading and SR-IOV has made negligible difference (some for obvious reasons such as dedicated nics to the VM plus low load on host).
Will also move Untangle to an appliance this afternoon too - just to rule the virtualisation out - its a Caswell i7 with 8GB ram and intel nics... will be interesting to see where it sits or indeed if anything changes!
The one thing that does puzzle me is why testing with multiple PCs and target speed test sites does the aggregate not equate to something at or around the 900ish? There cannot be a capacity issue at my local exchange already.... orders just opened up 5-6 weeks ago.
David
-
nperf is great. It has a weird quirk on my machine - it doesn’t see the availability of IPv6. That’s very weird since my iPadOS Safari prefers IPv6 anyway. For example speedtest.aa.net.uk often uses IPv6. I don’t know if their website even has an IPv6 address, didn’t check.
Don’t bother with the app. It seems to give worse results and takes forever as it does a whole load of additional tests involving performance if your browser and youtube video playing and all kinds of such nonsense that has nothing to do with your internet access link which is what it should be testing.
I can’t see how to restrict app or website to do just one test alone, eg upload, to save time.
-
nperf is great. It has a weird quirk on my machine - it doesn’t see the availability of IPv6. That’s very weird since my iPadOS Safari prefers IPv6 anyway. For example speedtest.aa.net.uk often uses IPv6. I don’t know if their website even has an IPv6 address, didn’t check.
I've be cautiously rolling out IPv6 to a limited number of clients on my LAN using a dedicated dual-stack VLAN. It didn't see my IPv6 either despite it working fine on AAISPs tester.
Don’t bother with the app. It seems to give worse results and takes forever as it does a whole load of additional tests involving performance if your browser and youtube video playing and all kinds of such nonsense that has nothing to do with your internet access link which is what it should be testing.
I can’t see how to restrict app or website to do just one test alone, eg upload, to save time.
Apps frustrate me in general, too often they provide an inferior experience to just using the website. The really annoying one for me is speedtest.net who nag for me to use their app on Windows because "it performs better on fast connections", but it absolutely doesn't.
I don't think I ever once got a good result on their Windows app, it always shows much slower than the browser.