Afrihost - Pure Fibre Feedback Thread Part 2

ronnie_94 said:
Hi @AfriNatic @Afrigirl

I am unable to connect to user.ip.afrihost.co.za when trying from another Afrihost connection but I can on mobile data.

I can confirm nothing changed on our side, could it be a issue on Afrihosts side?
Click to expand...

Morning

Should be fixed. Please test again. It was an Openserve interconnecting issue due to the migrations to the new BNGs
 
KinsZA said:
After a few hours on whatsapp of many tests it got escalated to openserve. Liquidtelecom are doing some very weird routing. Will see what happens there.



Edit: When you get a response in email support saying the reason that getting 3mbit upload internationally is "Because the packets have further to travel so your ping goes up" just leaves mind blown, especially since for that to be true for throughput requires 400 000km length fiber.
Click to expand...
Hi.

I've gone through the ticket and the troubleshooting completed so far. Although the traceroute to the affected international servers looks normal, I'd recommend performing an iperf3 test to investigate the reported international upload speeds further.

As this test measures the actual throughput between your connection and a remote server, it will help us determine whether the issue is specific to the international upload path.
 
Afrigirl said:
Hi.

I've gone through the ticket and the troubleshooting completed so far. Although the traceroute to the affected international servers looks normal, I'd recommend performing an iperf3 test to investigate the reported international upload speeds further.

As this test measures the actual throughput between your connection and a remote server, it will help us determine whether the issue is specific to the international upload path.
Click to expand...
Hi

I got the "Please be advised that Afrihost does not have control over traffic routed internationally. " line from support this morning, with no requests for anything else.

I'd rather just move this to DM's at this point and update the thread should a resolution occur. I'm not using Afrihost for my data connection at the moment (didn't have a choice) so my service is fine but if its to help you potentially help someone else then I'll run tests but need specifics.
 
@Afrigirl — I think I may be seeing the same international upload issue as @KinsZA.

I'm on Afrihost/Openserve in Raithby, Western Cape, 100/50 Mbps. I reported this to support on 11 September, but first-line support eventually told me that because the problem appears to be on the Liquid/international leg, Afrihost couldn't do anything about it.

The symptoms are very specific:

  • Local upload is normal: ~48 Mbps
  • International download is reasonably healthy: 30–37 Mbps single-stream from London/New York
  • International uploadis severely degraded:
    • London: 0.6–0.95 Mbps single-stream
    • New York: ~0.65 Mbps
    • AWS US-East: ~0.53 Mbps
    • 8 parallel streams to London: starts at ~15.6 Mbps but quickly settles at only ~5–6 Mbps
So this isn't simply international latency. The same connection can download internationally at 30–37 Mbps, while uploads collapse to below 1 Mbps on a single TCP connection. My domestic upstream is demonstrably capable of ~48 Mbps.

I've also tested latency while the bad upload is actually happening. There is 0% packet loss, gateway latency stays around 2.4 ms, and the Liquid London hop stays around 146 ms under load. In other words, there's no indication that my Openserve last mile or local network is becoming congested.

The interesting bit is the routing. Both my London and New York tests leave Afrihost and then use the same Liquid Telecom SA → London path. After London they diverge — London goes via LINX/Clouvider while New York continues via GTT — yet upload performance collapses to both destinations. That makes the common Afrihost → Liquid international egress look particularly suspicious.

This also sounds extremely similar to what KinsZA has been reporting here recently, including the very low international upload while local performance remains fine.

I've produced a fairly detailed diagnostic report with the raw throughput tests, traceroutes, loaded/idle latency tests and the exact commands used (attached).

Could you please check whether my case can be linked to the same investigation as KinsZA's, rather than opening another generic Openserve line fault?

My concern is specifically international outbound capacity/congestion/policing on the Afrihost → Liquid path, not a routing failure. A traceroute being clean is expected here — the problem only appears during sustained bulk upload.

My Ticket Number is PES-586-39304
 

Attachments

Last edited:
karis said:
@Afrigirl — I think I may be seeing the same international upload issue as @KinsZA.

I'm on Afrihost/Openserve in Raithby, Western Cape, 100/50 Mbps. I reported this to support on 11 September, but first-line support eventually told me that because the problem appears to be on the Liquid/international leg, Afrihost couldn't do anything about it.

The symptoms are very specific:

  • Local upload is normal: ~48 Mbps
  • International download is reasonably healthy: 30–37 Mbps single-stream from London/New York
  • International uploadis severely degraded:
    • London: 0.6–0.95 Mbps single-stream
    • New York: ~0.65 Mbps
    • AWS US-East: ~0.53 Mbps
    • 8 parallel streams to London: starts at ~15.6 Mbps but quickly settles at only ~5–6 Mbps
So this isn't simply international latency. The same connection can download internationally at 30–37 Mbps, while uploads collapse to below 1 Mbps on a single TCP connection. My domestic upstream is demonstrably capable of ~48 Mbps.

I've also tested latency while the bad upload is actually happening. There is 0% packet loss, gateway latency stays around 2.4 ms, and the Liquid London hop stays around 146 ms under load. In other words, there's no indication that my Openserve last mile or local network is becoming congested.

The interesting bit is the routing. Both my London and New York tests leave Afrihost and then use the same Liquid Telecom SA → London path. After London they diverge — London goes via LINX/Clouvider while New York continues via GTT — yet upload performance collapses to both destinations. That makes the common Afrihost → Liquid international egress look particularly suspicious.

This also sounds extremely similar to what KinsZA has been reporting here recently, including the very low international upload while local performance remains fine.

I've produced a fairly detailed diagnostic report with the raw throughput tests, traceroutes, loaded/idle latency tests and the exact commands used (attached).

Could you please check whether my case can be linked to the same investigation as KinsZA's, rather than opening another generic Openserve line fault?

My concern is specifically international outbound capacity/congestion/policing on the Afrihost → Liquid path, not a routing failure. A traceroute being clean is expected here — the problem only appears during sustained bulk upload.

My Ticket Number is PES-586-39304
Click to expand...
On a different note, and being in software dev, I wish MY clients do this kind of diagnostics when reporting issues, rather than "doesn't work" :D
 
karis said:
@Afrigirl — I think I may be seeing the same international upload issue as @KinsZA.

I'm on Afrihost/Openserve in Raithby, Western Cape, 100/50 Mbps. I reported this to support on 11 September, but first-line support eventually told me that because the problem appears to be on the Liquid/international leg, Afrihost couldn't do anything about it.

The symptoms are very specific:

  • Local upload is normal: ~48 Mbps
  • International download is reasonably healthy: 30–37 Mbps single-stream from London/New York
  • International uploadis severely degraded:
    • London: 0.6–0.95 Mbps single-stream
    • New York: ~0.65 Mbps
    • AWS US-East: ~0.53 Mbps
    • 8 parallel streams to London: starts at ~15.6 Mbps but quickly settles at only ~5–6 Mbps
So this isn't simply international latency. The same connection can download internationally at 30–37 Mbps, while uploads collapse to below 1 Mbps on a single TCP connection. My domestic upstream is demonstrably capable of ~48 Mbps.

I've also tested latency while the bad upload is actually happening. There is 0% packet loss, gateway latency stays around 2.4 ms, and the Liquid London hop stays around 146 ms under load. In other words, there's no indication that my Openserve last mile or local network is becoming congested.

The interesting bit is the routing. Both my London and New York tests leave Afrihost and then use the same Liquid Telecom SA → London path. After London they diverge — London goes via LINX/Clouvider while New York continues via GTT — yet upload performance collapses to both destinations. That makes the common Afrihost → Liquid international egress look particularly suspicious.

This also sounds extremely similar to what KinsZA has been reporting here recently, including the very low international upload while local performance remains fine.

I've produced a fairly detailed diagnostic report with the raw throughput tests, traceroutes, loaded/idle latency tests and the exact commands used (attached).

Could you please check whether my case can be linked to the same investigation as KinsZA's, rather than opening another generic Openserve line fault?

My concern is specifically international outbound capacity/congestion/policing on the Afrihost → Liquid path, not a routing failure. A traceroute being clean is expected here — the problem only appears during sustained bulk upload.

My Ticket Number is PES-586-39304
Click to expand...
Hi.

I have escalated this to the NOC team to check with our upstream provider.
 
karis said:
@Afrigirl — I think I may be seeing the same international upload issue as @KinsZA.

I'm on Afrihost/Openserve in Raithby, Western Cape, 100/50 Mbps. I reported this to support on 11 September, but first-line support eventually told me that because the problem appears to be on the Liquid/international leg, Afrihost couldn't do anything about it.

The symptoms are very specific:

  • Local upload is normal: ~48 Mbps
  • International download is reasonably healthy: 30–37 Mbps single-stream from London/New York
  • International uploadis severely degraded:
    • London: 0.6–0.95 Mbps single-stream
    • New York: ~0.65 Mbps
    • AWS US-East: ~0.53 Mbps
    • 8 parallel streams to London: starts at ~15.6 Mbps but quickly settles at only ~5–6 Mbps
So this isn't simply international latency. The same connection can download internationally at 30–37 Mbps, while uploads collapse to below 1 Mbps on a single TCP connection. My domestic upstream is demonstrably capable of ~48 Mbps.

I've also tested latency while the bad upload is actually happening. There is 0% packet loss, gateway latency stays around 2.4 ms, and the Liquid London hop stays around 146 ms under load. In other words, there's no indication that my Openserve last mile or local network is becoming congested.

The interesting bit is the routing. Both my London and New York tests leave Afrihost and then use the same Liquid Telecom SA → London path. After London they diverge — London goes via LINX/Clouvider while New York continues via GTT — yet upload performance collapses to both destinations. That makes the common Afrihost → Liquid international egress look particularly suspicious.

This also sounds extremely similar to what KinsZA has been reporting here recently, including the very low international upload while local performance remains fine.

I've produced a fairly detailed diagnostic report with the raw throughput tests, traceroutes, loaded/idle latency tests and the exact commands used (attached).

Could you please check whether my case can be linked to the same investigation as KinsZA's, rather than opening another generic Openserve line fault?

My concern is specifically international outbound capacity/congestion/policing on the Afrihost → Liquid path, not a routing failure. A traceroute being clean is expected here — the problem only appears during sustained bulk upload.

My Ticket Number is PES-586-39304
Click to expand...
Same thing, can confirm.

I gave back a huge amount of tests in support tickets, speed tests do Germany, London and New York etc... , @Afrigirl has my ticket number which will reflect all the tests.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X