Slow(ish) Speeds on Full Fibre 900

[-Richie-]

BT will only guarantee the speed to the hub, if you are receiving the speeds then BT are providing the contractually agreed service.

Downloading from 3rd party sources including from Europe adds other factors, peering/transit agreements, the hard drive used to download and other PC components etc all play a factor outside BT's control.
Not being able to download at full speeds does not indicate you're not getting the speeds advertised.

If you're being issued a deadlock letter, that means BT have not been able to reach agreement with you, so your complaint has been escalated as front line advisors can't send deadlock letters, this part confuses me because you say it wasn't escalated but it must of.

[kst]

>BT will only guarantee the speed to the hub, if you are receiving the speeds then BT are providing the contractually agreed service.

There is no point of having a 900mbit package if the only way to utilise it is on web based speedtests and speedtests which are likely internal to BT's network.

The package isn't sold as 900mbit internally, and up to 240mbit externally.

>Downloading from 3rd party sources including from Europe adds other factors, peering/transit agreement

I've tested in the UK, Germany, Netherlands, and Ireland.

OVH have public peering with LONIX and private peering with Telehouse (with lots of capacity) who BT also peers with.
I don't think AWS, Leaseweb, NFORCE, and Hetzner will be much different.

>the hard drive used

Tested, read is 2.6GB/s and write is 1.8GB/s.

>other PC components etc

All tested.

I have previously saturated a 600mbit line, with Virgin Media, the issue does not sit with my set up.

>Not being able to download at full speeds does not indicate you're not getting the speeds advertised.

Not being able to hit the minimum speed does.

[Checked2013]

It would be interesting to see if you've done any tests using a different pc, as that may double check things like wrong MTU or a browser issue (Firefox has been known in the past for slow multi-threaded downloads).

You say that it was fine on VM, but what works on VM might not work on everything else.

Also, we might be getting ultra high speed connections, but you may have to wait a bit for the rest of the Internet to catch up. For the moment in many cases a 900mb connection works best if there is multiple people using the connection at a time and that might be a good test also, by trying to download on multiple devices at the same time.

One last thing, make sure you don't have jumbo frames enabled, as that can also slow downloads.

[kst]

>It would be interesting to see if you've done any tests using a different pc

Two computers, both worked fine over VM.

>MTU

MTU is 1492.

>browser issue

Tried Firefox and Chrome.

>You say that it was fine on VM, but what works on VM might not work on everything else.

BT has enough capacity, peering, and transit to handle hitting the minimum guaranteed download speed.

>Also, we might be getting ultra high speed connections, but you may have to wait a bit for the rest of the Internet to catch up. 

Gbit, 2Gbit, and 10Gbit servers tested, in multiple regions, on multiple providers - I'm not even hitting the minimum guaranteed speed, which I would be happy with.

>works best if there is multiple people using the connection at a time and that might be a good test also, by trying to download on multiple devices at the same time.

I've tried multi-threaded downloads and hitting around the same speeds.

[Checked2013]

My point about another computer, is to see if default values work and not one you've used previously, just in case there has been settings you've changed.

Usually providers of servers quote total speed (meaning that it's shared between all their servers)

I also didn't mean try multi-threaded downloads, but try doing downloads on let's say 3 to 4 devices at the same time and see what the total speeds are.

[kst]

>My point about another computer, is to see if default values work and not one you've used previously, just in case there has been settings you've changed.

Nothing really has been changed, they're default Windows 10 installs with nothing fancy going on.

>Usually providers of servers quote total speed (meaning that it's shared between all their servers)

iperf3 run, UK server as the client, uploading to another server running iperf3:

---

---

root@some_host:~# iperf3 -c lon.speedtest.clouvider.net -p 5200
Connecting to host lon.speedtest.clouvider.net, port 5200
[ 5] local some_ip port 43680 connected to another_ip port 5200
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-1.00 sec 264 MBytes 2.21 Gbits/sec 2139 167 KBytes
[ 5] 1.00-2.00 sec 222 MBytes 1.86 Gbits/sec 3168 319 KBytes
[ 5] 2.00-3.00 sec 226 MBytes 1.89 Gbits/sec 3256 180 KBytes
[ 5] 3.00-4.00 sec 224 MBytes 1.88 Gbits/sec 3049 174 KBytes
[ 5] 4.00-5.00 sec 223 MBytes 1.87 Gbits/sec 3734 134 KBytes
[ 5] 5.00-6.00 sec 225 MBytes 1.89 Gbits/sec 3388 160 KBytes
[ 5] 6.00-7.00 sec 225 MBytes 1.88 Gbits/sec 3361 130 KBytes
[ 5] 7.00-8.00 sec 225 MBytes 1.88 Gbits/sec 3298 141 KBytes
[ 5] 8.00-9.00 sec 223 MBytes 1.87 Gbits/sec 3326 226 KBytes
[ 5] 9.00-10.00 sec 225 MBytes 1.89 Gbits/sec 3604 146 KBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 2.23 GBytes 1.91 Gbits/sec 32323 sender
[ 5] 0.00-10.00 sec 2.23 GBytes 1.91 Gbits/sec receiver

Ookla Speedtests from the UK server, using their CLI tool:

Download: 9399.89 Mbps (data used: 9.7 GB)
Upload: 1916.97 Mbps (data used: 2.3 GB)

Download: 8398.18 Mbps (data used: 9.3 GB)
Upload: 1938.68 Mbps (data used: 2.2 GB)

Download: 6436.37 Mbps (data used: 4.7 GB)
Upload: 1912.91 Mbps (data used: 2.3 GB)

Download: 6052.92 Mbps (data used: 5.8 GB)
Upload: 1887.84 Mbps (data used: 1.8 GB)

As we can see, the UK server is sending at around 1.91Gb/s.

I've tested speeds from the biggest cloud/dedicated server hosts in the EU region, using their UK data centres where possible, including AWS, and still cannot get anywhere near the minimum guaranteed speed.

This is a BT network issue, there's nothing more I can do.

[kst]

>My point about another computer, is to see if default values work and not one you've used previously, just in case there has been settings you've changed.

Nothing really has been changed, they're default Windows 10 installs with nothing fancy going on.

>Usually providers of servers quote total speed (meaning that it's shared between all their servers)

iperf3 run, UK server as the client, uploading to another server running iperf3:

---

---

root@some_host:~# iperf3 -c lon.speedtest.clouvider.net -p 5200
Connecting to host lon.speedtest.clouvider.net, port 5200
[ 5] local some_ip port 43680 connected to another_ip port 5200
[ ID] Interval Transfer Bitrate Retr Cwnd
[ 5] 0.00-1.00 sec 264 MBytes 2.21 Gbits/sec 2139 167 KBytes
[ 5] 1.00-2.00 sec 222 MBytes 1.86 Gbits/sec 3168 319 KBytes
[ 5] 2.00-3.00 sec 226 MBytes 1.89 Gbits/sec 3256 180 KBytes
[ 5] 3.00-4.00 sec 224 MBytes 1.88 Gbits/sec 3049 174 KBytes
[ 5] 4.00-5.00 sec 223 MBytes 1.87 Gbits/sec 3734 134 KBytes
[ 5] 5.00-6.00 sec 225 MBytes 1.89 Gbits/sec 3388 160 KBytes
[ 5] 6.00-7.00 sec 225 MBytes 1.88 Gbits/sec 3361 130 KBytes
[ 5] 7.00-8.00 sec 225 MBytes 1.88 Gbits/sec 3298 141 KBytes
[ 5] 8.00-9.00 sec 223 MBytes 1.87 Gbits/sec 3326 226 KBytes
[ 5] 9.00-10.00 sec 225 MBytes 1.89 Gbits/sec 3604 146 KBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.00 sec 2.23 GBytes 1.91 Gbits/sec 32323 sender
[ 5] 0.00-10.00 sec 2.23 GBytes 1.91 Gbits/sec receiver

Ookla Speedtests from the UK server, using their CLI tool:

Download: 9399.89 Mbps (data used: 9.7 GB)
Upload: 1916.97 Mbps (data used: 2.3 GB)

Download: 8398.18 Mbps (data used: 9.3 GB)
Upload: 1938.68 Mbps (data used: 2.2 GB)

Download: 6436.37 Mbps (data used: 4.7 GB)
Upload: 1912.91 Mbps (data used: 2.3 GB)

Download: 6052.92 Mbps (data used: 5.8 GB)
Upload: 1887.84 Mbps (data used: 1.8 GB)

As we can see, the UK server is sending at around 1.91Gb/s.

I've tested speeds from the biggest cloud/dedicated server hosts in the EU region, using their UK data centres where possible, including AWS, and still cannot get anywhere near the minimum guaranteed speed.

This is a BT network issue, there's nothing more I can do.

[Checked2013]

I think you seem to be missing the point and probably why you aren't getting anywhere with BT.

Never mind being able or not able to transfer from servers in data centres. But can you for instance download the latest NVIDIA drivers and can you then do it on multiple computers at the same time? The reason for this test is to determine if you have local congestion and it's the only way you would be able to get BT to listen as they would not be interested on 3rd party servers.

[kst]

Over two devices, I saw a peak combined speed of around 38MB/s, 8MB/s higher than the peak I've seen of 30MB/s on previous tests, if I remember correctly.

Still below the minimum guaranteed speed advertised.

[Checked2013]

You could post some screenshots of that and a ping to Google or something similar while you are doing the downloads?

[kst]

Hi @Checked2013,

Done some more testing tonight and narrowed it down to single thread performance.

Turns out aria2 doesn't work very well on Windows and my initial tests to rule that out, which did, can be considered worthless.

Single threaded:

This matches what I'm seeing when downloading using Firefox, Chrome, or using FTPS.

Multi-threaded:

One graph:

Speedtest files, 8 threads:

OVH -

Leaseweb -

NFORCE -

Hetzner -

Contrast those to the speeds I was seeing on my first post.
The consistency as seen in the graphs isn't great, but at least the speeds are much higher than what I was seeing on a single thread download.

So the question is, why is single threaded network performance on the BT network so bad?

I'll phone BT in the morning to have these notes added to my complaint and see if they want to kick it back to Openreach for further investigation, as they apparently run and manage BT's internal network...