Afrihost POC Network - Afrigreen Feedback

Status
Not open for further replies.
jmacc said:
tracert www.google.co.za should be fine
Click to expand...
Done.
Code: 
Microsoft Windows [Version 10.0.10240]
(c) 2015 Microsoft Corporation. All rights reserved.

C:\Users\Lara>tracert google.co.za

Tracing route to google.co.za [216.58.223.35]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2     *        *        *     Request timed out.
  3    46 ms    45 ms    46 ms  169-1-1-249.ip.afrihost.co.za [169.1.1.249]
  4    61 ms    49 ms    50 ms  169-1-1-250.ip.afrihost.co.za [169.1.1.250]
  5    67 ms    46 ms    45 ms  41.170.79.236
  6    44 ms    45 ms    45 ms  google.jb1.napafrica.net [196.46.25.166]
  7    49 ms    50 ms    50 ms  72.14.239.129
  8    46 ms    45 ms    45 ms  jnb01s08-in-f35.1e100.net [216.58.223.35]

Trace complete.
 
LaraC said:
Done.
Code: 
Microsoft Windows [Version 10.0.10240]
(c) 2015 Microsoft Corporation. All rights reserved.

C:\Users\Lara>tracert google.co.za

Tracing route to google.co.za [216.58.223.35]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2     *        *        *     Request timed out.
  3    46 ms    45 ms    46 ms  169-1-1-249.ip.afrihost.co.za [169.1.1.249]
  4    61 ms    49 ms    50 ms  169-1-1-250.ip.afrihost.co.za [169.1.1.250]
  5    67 ms    46 ms    45 ms  41.170.79.236
  6    44 ms    45 ms    45 ms  google.jb1.napafrica.net [196.46.25.166]
  7    49 ms    50 ms    50 ms  72.14.239.129
  8    46 ms    45 ms    45 ms  jnb01s08-in-f35.1e100.net [216.58.223.35]

Trace complete.
Click to expand...

Those stats are not too good. According to the article on "Detecting a congested ADSL exchange - Guide" here on MBB Anything above 25 is suspicious and above 100 is pretty much guaranteed to be a problem. Note: seeing 100+ on later hops is normal if its an international site
 
Last edited:
jmacc said:
Those stats are not too good. According to the article on "Detecting a congested ADSL exchange - Guide" here on MBB Anything above 25 is suspicious and above 100 is pretty much guaranteed to be a problem.
Click to expand...
Latency is high because I am routing via Johannesburg. AfriGreen doesn't have IPC in the south.
I'll do a trace to my local exchange.
 
Local test.
Code: 
Microsoft Windows [Version 10.0.10240]
(c) 2015 Microsoft Corporation. All rights reserved.

C:\Users\Lara>tracert bras.afrihost.com

Tracing route to bras.afrihost.com [155.239.255.250]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2     6 ms     6 ms     6 ms  155.239.255.250

Trace complete.
 
Dianysis said:
Yeah but why does it seem to affect mostly uncapped and business uncapped?
Click to expand...

It doesn't. If Afrigreen's demand outweighs the capacity the experience will be poor across the board for resource intense protocols.
Either way - it's something that'll be fixed in production.
 
jmacc said:
Those stats are not too good. According to the article on "Detecting a congested ADSL exchange - Guide" here on MBB Anything above 25 is suspicious and above 100 is pretty much guaranteed to be a problem. Note: seeing 100+ on later hops is normal if its an international site
Click to expand...

That's because there's only one POP - in JHB.
So latency in the South/ East will be about 30ms higher.
 
LaraC said:
Local test.
Code: 
Microsoft Windows [Version 10.0.10240]
(c) 2015 Microsoft Corporation. All rights reserved.

C:\Users\Lara>tracert bras.afrihost.com

Tracing route to bras.afrihost.com [155.239.255.250]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2     6 ms     6 ms     6 ms  155.239.255.250

Trace complete.
Click to expand...

Unfortunately this does not mean much. You have to test to a valid end point. This implies you need to hit a site you'll actually browse to.
 
AfriGuy said:
It doesn't. If Afrigreen's demand outweighs the capacity the experience will be poor across the board for resource intense protocols.
Either way - it's something that'll be fixed in production.
Click to expand...
But then why are most capped users (from what I am seeing - could be wrong) not having the p2p issue?
As you said - across the board. That is hardly across board.
And nothing against LataC - but why is performance so darn good all the time?
 
jmacc said:
Unfortunately this does not mean much. You have to test to a valid end point. This implies you need to hit a site you'll actually browse to.
Click to expand...

That destination is your closest exchange/ DSLAM.

jmacc said:
Interesting. This will affect the user experience.
Click to expand...

Correct. And that's been the case from day 1.
This is a proof of concept network, so it's pointless rolling out presence across the country. The concept worked, so we're rolling out a production network based off rules setup on Afrigreen.
 
Dianysis said:
But then why are most capped users (from what I am seeing - could be wrong) not having the p2p issue?
As you said - across the board. That is hardly across board.
And nothing against LataC - but why is performance so darn good all the time?
Click to expand...

We've seen mixed reports. If contention is there, it's there. It'll affect and bandwidth profile.
I've personally seen mixed results during this week. Usenet seems good and torrents aren't great, then visa versa the next night.

Again, the concept has proven successful - so demand based issues are being fixed in production.
 
Last edited:
Test at 20:00 as requested. Cape Town is a bit slower than usual.
291267735.png
291268489.png
 
AfriGuy said:
That destination is your closest exchange/ DSLAM.

Why then do you not do a tracert via the closest exchange to a valid site?


Correct. And that's been the case from day 1.
This is a proof of concept network, so it's pointless rolling out presence across the country. The concept worked, so we're rolling out a production network based off rules setup on Afrigreen.
Click to expand...

A POC is open to a limited number of testers. When the number of testers reach the point where the POC equipment cannot handle the testing, should the POC not end? You keep saying the concept worked. This implies that based on feedback and testing results, the POC is complete. Why then keep this thread going? Why ask us to provide feedback?
 
jmacc said:
A POC is open to a limited number of testers. When the number of testers reach the point where the POC equipment cannot handle the testing, should the POC not end? You keep saying the concept worked. This implies that based on feedback and testing results, the POC is complete. Why then keep this thread going? Why ask us to provide feedback?
Click to expand...

That's exactly what's happened though, we aren't loading more clients onto the network.
Feedback is always valued, it's foolish to just cut things short.
 
LaraC said:
Although latency is a bit higher it doesn't affect browsing or YouTube. Pages load quickly.
Click to expand...

It is all about tolerance. If you are already operating close to the edge it will not take much to push the user experience off the rails. Common sense really. The basics needs to be done correctly. If they repeat this in the new shiny production network we'll have a problem. Like they say - Anything above 25 is suspicious
 
jmacc said:
It is all about tolerance. If you are already operating close to the edge it will not take much to push the user experience off the rails. Common sense really. The basics needs to be done correctly. If they repeat this in the new shiny production network we'll have a problem. Like they say - Anything above 25 is suspicious
Click to expand...
I have never had problems with streaming YouTube. Even at high resolution.
 
Status
Not open for further replies.
Back
Top
Sign up to the MyBroadband newsletter
X