Afrihost CAPPED ADSL Feedback (MTN)

[Worskos]

The 3rd hop latency is consistently high, so I would say there is something consistent there, even though later results may vary, there is definitely an issue not ISP related for those results to be consistent. If it's congestion for example, resetting your account may give slightly better results just after the port is reset, but will eventually be poorer.

It's also important to actually change your username on your router and not test with PPPoE passthrough. When your router manages traffic it affects throughput more than when you passthrough directly so the results can be different in my experience.

Really?? This seems like an answer just to shrug me off.
While the 3rd hop is higher than you would like on both accounts, the AH account is still 102ms slower than the WA account, does that not indicate anything to you???

It's also important to actually change your username on your router and not test with PPPoE passthrough

I don't know how to change this

 

[NeoNarchy]

What I've seen is that since Telkom upgraded exchanges there are sometimes differences, particularly between SAIX and non-SAIX accounts. I tested last night and for some reason all my other ISP test accounts were failing to get any throughput on P2P above 10kBps, while a SAIX account ran at just over 1MBps. Load shedding happened, then lights came on and my line was maxxing again. We've also seen outages where no accounts will auth but Telkom accounts do, and I've also seen different ISP accounts route to different ESR routers at the BRAS, which produce vastly different results. Until we fully understand the changes made to the network (in terms of Telkom's infrastructure) the more info we can get the better.

Fair enough but i am using 3 MTN accounts (2 AH & 1 Axxess) & 1 IS account the 3 MTN accounts don't work like it should the IS(Webafrica) account does. This is how i feel http://imgur.com/qkJl0IP (had to redo my image i suck at paint editing)

 

[AfriMan]

Really?? This seems like an answer just to shrug me off.
While the 3rd hop is higher than you would like on both accounts, the AH account is still 102ms slower than the WA account, does that not indicate anything to you???

I don't know how to change this

I'm definitely not trying to shrug you off but just pointing out that this is not a definitive indicator. Consistent latency between both accounts would mean that the destination latency could be inconsistent, regardless of which account is used and it's more likely coincidence or related to which account was used closer to a port reset that determines the difference in performance.

So I would also recommend logging onto the router, changing username on that router (if you have the model number I can check my user manual directory) and then doing a comprehensive test.

 

[AfriMan]

Fair enough but i am using 3 MTN accounts (2 AH & 1 Axxess) & 1 IS account the 3 MTN accounts don't work like it should the IS(Webafrica) account does. This is how i feel http://imgur.com/qkJl0IP (had to redo my image i suck at paint editing)

I totally understand what you are saying here, and I know that we have some challenges on our side that we are addressing. But right now we can't be sure if outside factors are influencing your results.

 

[Worskos]

I'm definitely not trying to shrug you off but just pointing out that this is not a definitive indicator. Consistent latency between both accounts would mean that the destination latency could be inconsistent, regardless of which account is used and it's more likely coincidence or related to which account was used closer to a port reset that determines the difference in performance.

So I would also recommend logging onto the router, changing username on that router (if you have the model number I can check my user manual directory) and then doing a comprehensive test.

It's a D-Link 2750U
What about the PPPoE thing?

 

[NeoNarchy]

I totally understand what you are saying here, and I know that we have some challenges on our side that we are addressing. But right now we can't be sure if outside factors are influencing your results.

Define outside factors? AH & axxess are having issues but yet an alt account does not,where is the problem then?

 

[Zassie]

The 3rd hop latency is consistently high, so I would say there is something consistent there, even though later results may vary, there is definitely an issue not ISP related for those results to be consistent. If it's congestion for example, resetting your account may give slightly better results just after the port is reset, but will eventually be poorer.

It's also important to actually change your username on your router and not test with PPPoE passthrough. When your router manages traffic it affects throughput more than when you passthrough directly so the results can be different in my experience.

Is this not caused by the fact that telkom is blocking ICMP traffic on many of the exchanges now?
I can have rock solid ping to any IP (When afrihost isn't acting up) but as soon as i run a tracert its the same story...

1: router hop
2: timeout
3: huge spike
4: normalises for the rest of the hops.

I was informed by a friend that works at an ISP that telkom has started to block ICMP traffic at it's exchanged.

 

[nicki]

We definitely don't have protocol or port shaping on Capped or Business accounts. But if the network is showing high contention than optimal, you may see some services perform better than others until demand subsides.

How is that not shaping?

 

[Davedes]

We definitely don't have protocol or port shaping on Capped or Business accounts. But if the network is showing high contention than optimal, you may see some services perform better than others until demand subsides.

Absolutely but then all services would be effected and not just downloads which is what I'm experiencing.

 

[Worskos]

I'm definitely not trying to shrug you off but just pointing out that this is not a definitive indicator. Consistent latency between both accounts would mean that the destination latency could be inconsistent, regardless of which account is used and it's more likely coincidence or related to which account was used closer to a port reset that determines the difference in performance.

So I would also recommend logging onto the router, changing username on that router (if you have the model number I can check my user manual directory) and then doing a comprehensive test.

This was from AfriGenie

Timing out on the second-hop is not actually an indication of the issue, as Telkom has recently made changes to certain Exchanges so that they no longer respond to ICMP requests. This basically means that the Exchange will not appear on your Traceroute which is why you see * * * on the second hop.

Does this have any bearing on my issue?

 

[AfriMan]

Absolutely but then all services would be effected and not just downloads which is what I'm experiencing.

Not necessarily it depends on how those services draw bandwidth. We don't have a rule in place so differences in performance can only be due to higher contention for bandwidth.

 

[AfriMan]

How is that not shaping?

We are not actively setting which services perform better with a policy or ruleset. If we were, that would be shaping of sorts. This is determined by the contention on the network due to demand, so we can't predict the behaviour from our side.

 

[AfriMan]

Define outside factors? AH & axxess are having issues but yet an alt account does not,where is the problem then?

we don't know what those factors are, to be honest, but there does seem to be something odd happening. I'll be running some tests tonight on all the ISP accounts I have to compare performance and if there does seem to be a pattern I'll share my findings.

Just hope Eskom doesn't knee me in the face again tonight

 

[AfriMan]

It's a D-Link 2750U
What about the PPPoE thing?

PPPoE passthrough is basically when you setup a broadband connection on your desktop, not the router and while your regular account is active you dial through using another ISP account. I think you would know if you were doing this

 

[Worskos]

PPPoE passthrough is basically when you setup a broadband connection on your desktop, not the router and while your regular account is active you dial through using another ISP account. I think you would know if you were doing this

So what is our next course of action to remedy this situation?

 

[AfriMan]

So what is our next course of action to remedy this situation?

We'll need to troubleshoot the line and see if the latency is consistent across all ISPs, as it seems to be, and then log a fault with Telkom most likely to confirm if this is congestion or a line fault.

 

[ogchucky]

Is afrihost shaping capped account on p2p?? under 20kb/s down speed. Yet full speed on normal downloads and streaming. Very fishy

 

[Worskos]

We'll need to troubleshoot the line and see if the latency is consistent across all ISPs, as it seems to be, and then log a fault with Telkom most likely to confirm if this is congestion or a line fault.

Logged a fault with Telkom already.
They said it was a problem with the sync and they have since fixed it. According to them, all is fine now.

As best and as simply as you can, make me understand why the AH trace is 102ms worse off than the WA one...

The thing is, from what I see here on the forums, there is something wrong with the shaping rules affecting capped users, right? Am I not just seeing the effects of that on my line as well?

Also, does this have any bearing on what's happening?

Timing out on the second-hop is not actually an indication of the issue, as Telkom has recently made changes to certain Exchanges so that they no longer respond to ICMP requests. This basically means that the Exchange will not appear on your Traceroute which is why you see * * * on the second hop.

 

[Davedes]

Not necessarily it depends on how those services draw bandwidth. We don't have a rule in place so differences in performance can only be due to higher contention for bandwidth.

So streaming, YouTube, browsing, online gaming etc all work at full line speed of 8Mbps but downloads won't go past 30KB no matter what I do and & you expect me to believe it's due to contention.

Bullsht...

 

[AfriMan]

Logged a fault with Telkom already.
They said it was a problem with the sync and they have since fixed it. According to them, all is fine now.

As best and as simply as you can, make me understand why the AH trace is 102ms worse off than the WA one...

The thing is, from what I see here on the forums, there is something wrong with the shaping rules affecting capped users, right? Am I not just seeing the effects of that on my line as well?

Also, does this have any bearing on what's happening?

It is possible that our current demand (during peak) could be worsening the latency, but from the alternative ISP test you ran there is certainly latency on your line either way.

Remember that it's not our rules affecting capped users, but the overall increase in demand. We're trying to implement rules to shape rogue P2P traffic on Uncapped accounts, but we're not using shaping rules to affect Capped users in any way.