Afrihost CAPPED ADSL Feedback (MTN)

[Timber_MG]

Just because the jitter and latency appears on nodes further upstream doesn't mean that the issue is directly there. Telia connections likewise had the same issues. IS went onto L3 and Telia just fine with very low latency. 100ms jitter on your intl transit paints a very clear picture.

 

[Beri]

What is strange to me is that AH tells me that anything below 100ms is great and that my average of 50ms is normal on the first couple of hops. Yet I see in a lot of the post pings of under 10ms.....I have never seen this. only from computer to router.
I get timed out of the second hop and this was escalated to Telkom and nothing was done.
I am not expert in any on this but it seems that AH takes advantage of this and feeds me BS.
AH contacted Telkom to look into the problem. They should tell Telkom what the problem is and how to fix it.

I know my son keeps on complaining of high pings when playing online.

Last ref is AH20140703-132080

 

[Timber_MG]

Have your son use a free WA account and set up a PPPoE connection with your router in half bridge mode if possible. His pings will improve but more importantly the sporadic packet loss and jitter (when the pipes that feed you get congested and some packets arrive later by some margin) isn't an issue meaning less rubber banding and glitching.

 

[Timber_MG]

@Beri, if you have a lot of latency to the first two hops, you have a problem with the exchange or the line. In my case I can see when afrihost's limits starts to come in at hop 3 onwards when i compare it to IS with jitter often in the 15-20ms range (jitter is +/-) compared to 3-4ms on IS.

 

[AfriMan]

Just because the jitter and latency appears on nodes further upstream doesn't mean that the issue is directly there. Telia connections likewise had the same issues. IS went onto L3 and Telia just fine with very low latency. 100ms jitter on your intl transit paints a very clear picture.

By that reason, the tests aren't really conclusive. Again, we'd need to see this consistently across the entire network, and it doesn't show on our testing internally. Throughput and latency are testing great to the EU.

 

[AfriMan]

What is strange to me is that AH tells me that anything below 100ms is great and that my average of 50ms is normal on the first couple of hops. Yet I see in a lot of the post pings of under 10ms.....I have never seen this. only from computer to router.
I get timed out of the second hop and this was escalated to Telkom and nothing was done.
I am not expert in any on this but it seems that AH takes advantage of this and feeds me BS.
AH contacted Telkom to look into the problem. They should tell Telkom what the problem is and how to fix it.

I know my son keeps on complaining of high pings when playing online.

Last ref is AH20140703-132080

My understanding is that for the second hop (which is generally the gateway to our network) latency should be below 30ms, but should be around 10ms or less for best performance. If it's as high as 50ms, then there is an issue of concern there.

With regard to the 2nd hop issue, that was definitely on Telkom's side, confirmed by another client as well. We can't force Telkom to fix it or dictate to them. Their response was that it doesn't interfere with performance on the line.

 

[Timber_MG]

By that reason, the tests aren't really conclusive. Again, we'd need to see this consistently across the entire network, and it doesn't show on our testing internally. Throughput and latency are testing great to the EU.

What is conclusive however is that when I change to an IS account the experience improves drastically.
If your internal testing doesn't replicate the end-user experience, you should look at changing the guys who do the product testing. Simple QA.

I'm not the only one who had terrible international lately. Since testing cybersmart and IS it is abundantly clear that for EU content Afrihost is last of the class for some time now, especially where latency is of any concern. Lately it's been really poor and variable. Only late last night did the latency settle about 10-20ms higher than the weekend and last week.

For all I know your "testing" changes to the shapers are having an influence on traffic higher up the distribution tree or you're perhaps not managing traffic at the peering points properly and just hopping on hoping that your management further downstream is sufficient. If you were more transparent us users wouldn't have to second guess what Afrihost is doing next.

If you're not reproducing these issues, perhaps you should rent some DSL lines and do proper end user experience testing. Switch PPPoE accounts on a router on some DSL device remotely and test instead of just testing off your network. Is Afrihost pushing the envelope of its business case again, speculation is rife of you tweaking pre take-over?

 

[AfriMan]

What is conclusive however is that when I change to an IS account the experience improves drastically.
If your internal testing doesn't replicate the end-user experience, you should look at changing the guys who do the product testing. Simple QA.

I'm not the only one who had terrible international lately. Since testing cybersmart and IS it is abundantly clear that for EU content Afrihost is last of the class for some time now, especially where latency is of any concern. Lately it's been really poor and variable. Only late last night did the latency settle about 10-20ms higher than the weekend and last week.

For all I know your "testing" changes to the shapers are having an influence on traffic higher up the distribution tree or you're perhaps not managing traffic at the peering points properly and just hopping on hoping that your management further downstream is sufficient. If you were more transparent us users wouldn't have to second guess what Afrihost is doing next.

If you're not reproducing these issues, perhaps you should rent some DSL lines and do proper end user experience testing. Switch PPPoE accounts on a router on some DSL device remotely and test instead of just testing off your network. Is Afrihost pushing the envelope of its business case again, speculation is rife of you tweaking pre take-over?

There is a lot of speculation here, and I think there is very little basis for this. Ultimately, we are happy with our QA process. We test with normal Telkom lines, to the same services and account policies that clients test with. Again, we can speak with interactions with tens of thousands of clients in terms of tests run with them, overall experience and specific trends and commonalities that we can marry with specific targeted replicated results in controlled and real-world conditions. We've been doing this for a long time, MTN's team even longer, and I think we do know how to apply best practice as well as common sense.

Ultimately, it much more productive for clients to report on their own experience, and we deal with the individual information and circumstances specific to that client, rather than assume that each client's experience is general to the whole network. It means we'll find solutions a lot quicker.

 

[Timber_MG]

The argument that your "internal testing" is at odds with user experience holds. If you claim your testing (by extension to Telkom lines reflecting actual user experience) doesn't show up issues experienced by the community, what gives?

What's going on with intl, please give us a response that's got at least some basic network terminology and not some slippery PR.

 

[roadcat]

My understanding is that for the second hop (which is generally the gateway to our network) latency should be below 30ms, but should be around 10ms or less for best performance. If it's as high as 50ms, then there is an issue of concern there.

With regard to the 2nd hop issue, that was definitely on Telkom's side, confirmed by another client as well. We can't force Telkom to fix it or dictate to them. Their response was that it doesn't interfere with performance on the line.

The 2nd hop timeout is apparently because of new ESRs used in the MSAN and is expected behaviour. What is not expected is that it might be responsible for messing up the routing and I get increased latency as a result.

Using Plugg I return to sub 20ms latency. With AH this is 50ms (routes via JHB). In the past my AH latency was also sub 20ms.

 

[sgs]

The argument that your "internal testing" is at odds with user experience holds. If you claim your testing (by extension to Telkom lines reflecting actual user experience) doesn't show up issues experienced by the community, what gives?

What's going on with intl, please give us a response that's got at least some basic network terminology and not some slippery PR.

+1

 

[VScheepers]

+1

+1 here too....

and suddenly the PR has gone quiet.

 

[AfriMan]

The argument that your "internal testing" is at odds with user experience holds. If you claim your testing (by extension to Telkom lines reflecting actual user experience) doesn't show up issues experienced by the community, what gives?

What's going on with intl, please give us a response that's got at least some basic network terminology and not some slippery PR.

I've been pretty straightforward here. If you wanna call that "slippery PR", I guess you're entitled to your opinion.

 

[AfriMan]

The 2nd hop timeout is apparently because of new ESRs used in the MSAN and is expected behaviour. What is not expected is that it might be responsible for messing up the routing and I get increased latency as a result.

Using Plugg I return to sub 20ms latency. With AH this is 50ms (routes via JHB). In the past my AH latency was also sub 20ms.

It's quite possible that the routing would add on additional latency. Just not sure how to further address this, cos on our side we're routing you in the right place, but we're not sure how you're ending up at the wrong place

 

[Timber_MG]

I've been pretty straightforward here. If you wanna call that "slippery PR", I guess you're entitled to your opinion.

Slippery PR is claiming to be transparent but not providing any relevant detail whatsoever or denying user experience with statements like internal testing shows up everything is OK/normal atm.

 

[AfriMan]

Slippery PR is claiming to be transparent but not providing any relevant detail whatsoever or denying user experience with statements like internal testing shows up everything is OK/normal atm.

That is not everything that I mentioned, I also mentioned that we have information and data via other channels and through our call centre where we do extensive testing and troubleshooting with clients.

I've never denied what you are experiencing individually, I've said that your sweeping statements regarding the state of the network (based on your own individual experience) are not consistent with the general reports we receive across tens of thousands of clients and our own internal tests and monitoring.

 

[Timber_MG]

I am not basing my comments on my experience alone. If you claim it's an individual issue but others have similar experiences on this forum then I do take issue.

My extrapolations are based off my own experience as well as feedback on this forum and others who use your network. The access mechanism between IS and yourselves is very similar, meaning that even though I am replicating the issue successfully here, changing to another provider that also uses IPC and exactly the same leaf node for access narrows down matters considerably.

I go further and state the local content is accessed significantly better than international at the moment, moving emphasis to your international transit. Comparing with IS for example highlights this quite sufficiently.

Direct questions:

Why is international giving either packet loss, jitter or hugely increased latency over the last few weeks and why has this been intermittent?

Have you replicated the issues that are appearing on this site and others?

Have you identified potential solutions and can you communicate time frames for resolution or whether no change is foreseen?

 

[AfriMan]

I am not basing my comments on my experience alone. If you claim it's an individual issue but others have similar experiences on this forum then I do take issue.

My extrapolations are based off my own experience as well as feedback on this forum and others who use your network. The access mechanism between IS and yourselves is very similar, meaning that even though I am replicating the issue successfully here, changing to another provider that also uses IPC and exactly the same leaf node for access narrows down matters considerably.

I go further and state the local content is accessed significantly better than international at the moment, moving emphasis to your international transit. Comparing with IS for example highlights this quite sufficiently.

Direct questions:

Why is international giving either packet loss, jitter or hugely increased latency over the last few weeks and why has this been intermittent?

Have you replicated the issues that are appearing on this site and others?

Have you identified potential solutions and can you communicate time frames for resolution or whether no change is foreseen?

So as mentioned, you are extrapolating, which is theoretical. We have actual experience and feedback along with hard results from tests with clients, internal tests as well as constant monitoring from various points within the network between ourselves and MTN. So we're not guessing on our side.

We have not replicated significant packet loss or jitter on our side or hugely increased latency either to local or EU based sites. We have seen throughput to the US is at times less than we'd like, but latency appears to be consistent in feedback and our tests.

Until we can identify actual problems, which are mainly related to US throughput, timeframes would not be realistic.

 

[zll]

Hi Afriman,

I am currently getting huge ping spikes on my capped account, and had almost unplayable ping last night (400 to EU last night, anywhere between 180 and 350 this evening). I have tested another IS based service provider and it's rock solid so it's safe to assume it's not my side. Are you guys currently have any issues?

 

[AfriGuy]

Hi Afriman,

I am currently getting huge ping spikes on my capped account, and had almost unplayable ping last night (400 to EU last night, anywhere between 180 and 350 this evening). I have tested another IS based service provider and it's rock solid so it's safe to assume it's not my side. Are you guys currently have any issues?

There shouldn't be anything causing this from our end, do you perhaps have a few comparative traceroutes I can look at? I'd like to see where the additional latency is coming in.