Slow FTTP: Difference between revisions
m (→example) |
m (→example) |
||
Line 52: | Line 52: | ||
Here is an example where we are running a single thread iperf3 test for 10 minutes, and then a 10-thread test for 10 minutes - the difference is shown clearly in our CQM graphs: |
Here is an example where we are running a single thread iperf3 test for 10 minutes, and then a 10-thread test for 10 minutes - the difference is shown clearly in our CQM graphs: |
||
[[File:Fttp-single-thread-slow. |
[[File:Fttp-single-thread-slow.png|frame|none|Alternating between single and multi-thread iperf download tests]] |
||
== I may have packet loss == |
== I may have packet loss == |
Revision as of 10:01, 14 Haziran 2023
There are a few things that could cause a FTTP circuit to be slower than expected.
- A fault somewhere causing packetloss (this could be local to the customer's home network, or something further afield)
- Some miss-configuration or fault in the fibre provider's network (eg broken links causing congestion)
Here is some steps to take if you have problems with your FTTP service.
I have a slow FTTP connection
If your FTTP service is slow, eg a speed test is not reporting expected speeds, then try these:
- Use our speed test and some other 3rd party ones to check up and download speeds: https://speedtest.aa.net.uk https://www.fast.com- try tests at different times of the day as this can help track 'peak time congestion' - note - many http speed testers are not very good for testing 1G connections - to always take results with a pinch of salt
- Try a wired connection from your computer instead of WiFi to rule out any WiFi problems.
- Check/swap the cable between the ONT (The Openreach/CityFibre unit that the fibre connects to) and your router - ensure it's a 8-wire CAT5 or CAT6 cable - not one with 4 wires as these will run at 100M and not 1G.
- Ensure the router is capable of the speed of the circuit - eg, maybe you have a 1G service but an older router? - let us know the model you are using
- Check the CQM Graphs, via the Control Pages, to see if it is slow due to lots of traffic.
- Ask us about testing with 'iperf' - as this is generally better than http speed test websites. See: https://support.aa.net.uk/Windows_iperf3
- Try a laptop/PC plugged directly in to the ONT instead of using your router - set up a new network connection of type PPPoE and use your xxx@a login and password to connect, and run a speed test. see: PPPoE_on_a_Computer for more help on this.
- Try a different computer/device
- Try a power cycle of the Openreach/CityFibre ONT
- Check for packet loss - see below.
- ...get in touch.
Always worth running a test, even if LOS light is off - there could be a problem further upstream and Openreach may already be aware or we'll need to report a fault:
- Run a test:-
- Our Control Page will allow you to run an 'End to End' test, this may if Openreach's systems detect a fault, in which case, do get in touch. eg: GTC_FTTP_SERVICE_1005 Possible fault in the Openreach network.
Single thread Vs multi thread performance problems
Sometimes a speed problem occurs for single-thread transfers, whist multi-thread transfers are OK. Broadly speaking, single thread Vs multi-thread speeds should be very similar, and near the top end of the speed of your circuit.
These can be tricky to diagnose and fix, as sometimes the problem is down to congestion in the back-haul network.
As well as testing the usual things (Different computer/router/directly-wired etc) to help diagnose, we suggest using iperf3 rather than http based speed tests as iperf3 is usually more reliable.
Support staff can give access to our own iperf server, and then you can run tests such as:
Single thread transfer from us to you (testing your download)
iperf3 -R -c <speedtest server>
10-thread transfer from us to you (testing your download)
iperf3 -R -P 10 -c <speedtest server>
Single thread transfer from you to us (testing your upload)
iperf3 -c <speedtest server>
10-thread transfer from you to us (testing your upload)
iperf3 -P 10 -c <speedtest server>
Running these at different times of day will give a picture as to whether single-thread is a problem, and if it's a problem due to the time of the day.
example
Here is an example where we are running a single thread iperf3 test for 10 minutes, and then a 10-thread test for 10 minutes - the difference is shown clearly in our CQM graphs:
I may have packet loss
Packetloss can be seen in a number of ways, check the following to see if the location can be narrowed down to something specific:
- Ping your router's LAN IP - this will be the 'default gateway' your computers use - eg, could be something like 192.168.1.1 - if you have loss pinging your local router, check with a wired connection direct in to the router...
- Check the CQM Graphs on the A&A Control pages
- Ping the A&A endpoint IP:
ping 81.187.81.187
- ping somewhere on the internet, eg
ping 8.8.8.8
andping bbc.co.uk
Get in touch with us about what you find.
Wrongly configured OLT
Packet loss with an odd pattern like this was seen after a CityFibre OLT was upgraded in March 2023 and apparently have some wrong configuration and caused around 5% loss:
Slow CityFibre since migrating
We have seen CityFibre connections limited to say 500M when they have been migrated over to us. This could be due to the previous service being a 500M service and CityFibre's systems not updating the profile/bandwidth for the circuit. This can be tested with connecting a PC to the ONT and setting PPPoE_on_a_Computer - if around 500M is the maximum then let us know. The fix is for CityFibre to re-provision the ONT, which we can request.