Afrihost CAPPED ADSL Feedback (MTN)

Status
Not open for further replies.
We've managed to isolate why demand appeared to peak and that seems to be under control now. But in terms of packet loss, we need to see test results showing exactly where in the routing packet loss occurs. In game reports don't tell us enough. For us to take action, it needs to be on a router or server within our network.

The problem is I don't think I can give that to you. Every time I've been getting loss in game and in mumble and then done pings/tracerts to that exact server there is no loss. It's only ever in game/application that loss occurs.
 
A ping uses a different protocol than what the game would use, which could explain why you don't see loss on a ping but you do in the game. The best would be to use something like wireshark to capture which ports/protocols and ips are being used.
 
Hi AfriTeam

Just a quick question about the ClientZone...

I have two different results for data used this month in my CZ...

CCZ1.jpg

CZ2.jpg

I am probably being doff and missing something glaringly obvious... but just in case... thought I'd hola at ya'all.
 
The problem is I don't think I can give that to you. Every time I've been getting loss in game and in mumble and then done pings/tracerts to that exact server there is no loss. It's only ever in game/application that loss occurs.

You should be able to get the mumble server's IP and then measure packet loss to that IP. If it's only in-game and doesn't show up on any test, how the game server is managing traffic or perhaps combining P2P elements into gameplay traffic.
 
A ping uses a different protocol than what the game would use, which could explain why you don't see loss on a ping but you do in the game. The best would be to use something like wireshark to capture which ports/protocols and ips are being used.

A packet is a packet though, it shouldn't really matter what the traffic classification is if a router or network switch is dropping packets. At least that would be my understanding :(
 
Hi AfriTeam

Just a quick question about the ClientZone...

I have two different results for data used this month in my CZ...

View attachment 198364

View attachment 198366

I am probably being doff and missing something glaringly obvious... but just in case... thought I'd hola at ya'all.

If you did a topup last month, it could be that your previous month's topup is still being reported in the current month's data allocation.

Would have to check for you if that's not the case :(
 
You should be able to get the mumble server's IP and then measure packet loss to that IP. If it's only in-game and doesn't show up on any test, how the game server is managing traffic or perhaps combining P2P elements into gameplay traffic.

Even if it were combining p2p elements, if we are using a capped account, why would it matter? Torrents aren't being affected at all on my side - full 4MB/s.

Edit - What is making me most upset is the feedback from you guys. It seems as though your team isn't pulling out all he stops to get this fixed ASAP. All I'm seeing are time-buying responses. That's just the vibe I'm getting from you. Forgive me if I'm seeing it wrong.
 
Last edited:
A packet is a packet though, it shouldn't really matter what the traffic classification is if a router or network switch is dropping packets. At least that would be my understanding :(
That is correct, when there is not a shaping device in play. Afrihost, or MTN atleast is making use of devices from Allot from what I last heard. To do traffic shaping, those devices actually drop packets.

My guess as to what is happening:
1. Afrihost/MTN give high priority to ICMP, so that traceroutes and pings look great. <-- This is good.
2. Afrihost/MTN give different priorities to different protocols, to give the best possible experience to customers. <-- This is good.
3. Afrihost/MTN give low or best-effort priority to "unknown" protocols. In Allot speak, this is called "Other UDP" and "Other TCP". <-- This is where the problem come into play.

There are 2 ways to fix this.
1. Unthrottle "Other TCP" and "Other UDP". I bet all these packet loss issues will disappear. The problem with this is, it will also unthrottle encrypted torrents and P2P.
or
2. Make a host list of the game and teamspeak servers that your customers use, give those high priority. The problem with this is, it will be annoying to manage.

You can choose which method you use, but thats pretty much what it boils down to.

Here is how the packet loss looks on an Afrihost account on my DSL line at home:

Screen Shot 2015-03-11 at 1.38.27 PM.jpg



And here is what it looks like on a Cybersmart account:

Screen Shot 2015-03-11 at 1.41.30 PM.jpg
 
Last edited:
If you did a topup last month, it could be that your previous month's topup is still being reported in the current month's data allocation.

Would have to check for you if that's not the case :(

Only mobile data top-up last month... ADSL not.

I am looking at it correctly I hope... and not wasting your time...
 
That is correct, when there is not shaping device in play. Afrihost, or MTN atleast is making use of devices from Allot from what I last heard. To do traffic shaping, those devices actually drop packets.

My guess as to what is happening:
1. Afrihost/MTN give high priority to ICMP, so that traceroutes and pings look great. <-- This is good.
2. Afrihost/MTN give different priorities to different protocols, to give the best possible experience to customers. <-- This is good.
3. Afrihost/MTN give low or best-effort priority to "unknown" protocols. In Allot speak, this is called "Other UDP" and "Other TCP". <-- This is where the problem come into play.

There are 2 ways to fix this.
1. Unthrottle "Other TCP" and "Other UDP". I bet all these packet loss issues will disappear. The problem with this is, it will also unthrottle encrypted torrents and P2P.
or
2. Make a host list of the game and teamspeak servers that your customers use, give those high priority. The problem with this is, it will be annoying to manage.

You can choose which method you use, but thats pretty much what it boils down to.

Here is how the packet loss looks on an Afrihost account on my DSL line at home:

View attachment 198386



And here is what it looks like on a Cybersmart account:

View attachment 198390

In broad strokes this is correct, we would give priority according to traffic. However, we do regular updates on traffic from TS, Mumble and any other services which may be otherwise classified but are known services that we don't want to shape. This is an ongoing process with our local and overseas providers and updated, along with regular batch updates we receive from the hardware/software vendors.

So it would not make sense if a service which was previously performing at peak suddenly began experiencing high packet loss unless either the traffic signature for the service changed (as we saw with Xbox traffic moving more towards P2P) or the packet loss occurs elsewhere. I would understand if packet loss occurred on these services if the network was saturated due to load on resources, but that doesn't appear to be the case here, so we need to understand what may have changed or what is being done differently on either end.
 
You should be able to get the mumble server's IP and then measure packet loss to that IP. If it's only in-game and doesn't show up on any test, how the game server is managing traffic or perhaps combining P2P elements into gameplay traffic.

As I said tracerts and ping tests do not have the same loss as in game and in application.

Just gave it a quick go and was getting 0% loss on the jhb server and 0 - 3% on the cape town one however they were both pretty empty. I'll try again at 4pm when I started yesterday for the sake of consistency.
 
As I said tracerts and ping tests do not have the same loss as in game and in application.

Just gave it a quick go and was getting 0% loss on the jhb server and 0 - 3% on the cape town one however they were both pretty empty. I'll try again at 4pm when I started yesterday for the sake of consistency.

Yes, let's see what the tests show when you are seeing packet loss in game.
 
In broad strokes this is correct, we would give priority according to traffic. However, we do regular updates on traffic from TS, Mumble and any other services which may be otherwise classified but are known services that we don't want to shape. This is an ongoing process with our local and overseas providers and updated, along with regular batch updates we receive from the hardware/software vendors.

So it would not make sense if a service which was previously performing at peak suddenly began experiencing high packet loss unless either the traffic signature for the service changed (as we saw with Xbox traffic moving more towards P2P) or the packet loss occurs elsewhere. I would understand if packet loss occurred on these services if the network was saturated due to load on resources, but that doesn't appear to be the case here, so we need to understand what may have changed or what is being done differently on either end.
I get what you say, and that is true. Right now however, something definitely is not right, as you can see those graphs in my previous post speak wonders about the packet loss on Afrihost, vs on Cybersmart where there is almost none.

If your IPCs are not at capacity, its fairly simple, it is the shaping devices that is throttling the traffic, so a normal MRTG graph will show you have loads of capacity available, when in reality the shaping device is actually artificially making it look lower than what the actual demand for traffic is. That would very well explain why there is so much packet loss issues right now.

Just look at those latency/packet loss graphs in my previous post again, and tell me if that is something you are proud off, knowing it is red on afrihost, and green on cybersmart. Same dsl line, same exchange, same routers, only different PPPoE sessions to different ISPs, both capped accounts.
 
Also I just want to add quick this is NOT something new. I've been having this loss for a while now. It certainly did not start this month or even last.
 
Just want to chime in and say that the entire MTN business ADSL network in the South has an inherent packet loss issue. It's one of the main reasons I left Afrihost.

Both Axxess and Afrihost show the same consistent packet loss on my end. I thought they would have figured this out by now though.
 
Just want to chime in and say that the entire MTN business ADSL network in the South has an inherent packet loss issue. It's one of the main reasons I left Afrihost.

Both Axxess and Afrihost show the same consistent packet loss on my end. I thought they would have figured this out by now though.

+1 During Gaming, I find that Afrihost is the worst.

The unidentifiable issue with packet loss on Heroes of the Storm as well as League of Legends now again is rather frustrating.
 
Status
Not open for further replies.
Top
Sign up to the MyBroadband newsletter
X