LATENCY!!!

While I appreciate that MWeb are pressing to get this issue resolved (and I appreciate the feedback from Will@MWeb), the fact remains that a week ago I was on the SAIX network, and flying!

The same week that MWeb have moved all their customers back to their own network, has seen severe congestion on the loop between the local exchanges and the MWeb Cape Town IPC. Conspiracy theorists may see some 'dark-forces' at work here...

This is getting to the point where I regret having an MWeb all inclusive package, since I can't switch to an SAIX based provider when these all to common hiccups occur...
 
Hi Gordon

Conspiracy theories aside for a moment, I'm not entirely sure I'm following you here:

When we were load balancing some of our users onto the SAIX network up until about 2 weeks ago the latency issues were present on both our IPC network and the SAIX based connections. This is in keeping with some of the other posts in this thread from Web Africa and Axxess users as well.

it's a bit difficult to explain this without a whiteboard handy and I'm not a network engineer myself, but as I understand it the various exchanges link up to a series of ESR (edge service routers). Our IPC links 'breakout' from the same network that these ESRs connect onto which allows us to route your traffic over our network as opposed to SAIX.

Now bearing that in mind, the two key facts that tell me the latency occurs at the level prior to reaching our IPC links are that:

a) Users on other ISPS are experiencing the same issues.
b) Tests run from a different location at the same time show much lower latency over the same routes (meaning that the latency must be coming from the interim steps not showing up on the trace).

My current theory on this, and I must stress that this is a theory which has not been borne out by feedback from Telkom as yet, is that there is an ESR servicing the Southern Suburbs area which is congested and that this is the underlying cause for the latency we are experiencing.

Kind Regards
Will
 
Will, you may be perfectly correct, but I have this feeling (may be wrong) that SAIX manages to route their own traffic differently than that they do for other ISPs (to the IPC step). Or perhaps the IPC and 10Mbps issues have been confused in my head (as well as by others), since they occurred around the same time.

Are other users (outside of Cape Town) experiencing similar issues? There are certainly some nasty comments on other threads. Does the CBD also go through the same route?
http://mybroadband.co.za/vb/showthr...-the-day-in-Cape-Town-CBD-since-10meg-upgrade
 
It's not an MWEB issue only, but a Telkom infrastructure issue. Lots of people across various ISPs are having problems :(
 
Gordon_R , as far as i know the Telkom accounts are stuffed too, I tested one when this started for me about 2 months ago.
 
Will, you may be perfectly correct, but I have this feeling (may be wrong) that SAIX manages to route their own traffic differently than that they do for other ISPs (to the IPC step). Or perhaps the IPC and 10Mbps issues have been confused in my head (as well as by others), since they occurred around the same time.

Are other users (outside of Cape Town) experiencing similar issues? There are certainly some nasty comments on other threads. Does the CBD also go through the same route?
http://mybroadband.co.za/vb/showthr...-the-day-in-Cape-Town-CBD-since-10meg-upgrade

Every ADSL customer regardless of ISP is routed via one of the Edge Service Routers. If you look at some of the other ISPS, then for example Internet Solutions and WEB Africa also have IPC based networks. I don't have a clear picture as to how Pure SAIX users are routed from the ESR onwards, but they would also certainly be affected by problems up to that level.

In terms of other areas that are affected it's a question of gathering the evidence and then raising the awareness.

I can tell you that we've escalated and had resolution of more than case where DSLAM backhaul links were over-utilized and needed to be upgraded. It seems that the Southern Suburbs issue is at a level further upstream and the feedback on this one is certainly going to be very interesting.

As I've mentioned to a previous poster in this thread if we can produce the troubleshooting results showing that there is an issue affecting multiple users then I'm happy to do what I can to take the matter forward.
 
Well, I do appreciate the feedback, but since this whole 10Mbps upgrade started, things seem to have gone backwards for an extended period of time...

It seems that no matter how much international and IPC capacity is installed, the loop from the DSLAM to the IPC is still being 'strangled' by Telkom infrastructure at present...

At peak times it takes as long for tracert to get from my DSLAM to the ESR (15km), than to get from the ESR to the UK via SEACOM (15,000km)!?
 
I feel your frustration Gordon. If I'm keeping an eye on things from home I have to call in to the office to have traces run from our test bench because I can't trust any of my own results.
 
Bad Latency's not only in the southern suburbs !

Hey guys its 20:07 10/10/2010 and I just did a ping test..... How horrible!

Pinging 168.209.1.140 with 32 bytes of data:
Reply from 168.209.1.140: bytes=32 time=187ms TTL=250
Reply from 168.209.1.140: bytes=32 time=171ms TTL=250
Reply from 168.209.1.140: bytes=32 time=197ms TTL=250
Reply from 168.209.1.140: bytes=32 time=194ms TTL=250

Back to 200ms odd.

I'm based in the Table View area. Seems that its not just a southern suburbs thing :/
Cant play any online games with this latency.

have a look here. Speed test to Newlands Server
http://www.pingtest.net/result/25787344.png

ADSL capped at 5mb/s with the Telkom Do broadband package.

ADDED * : I just did a speed test again and now all of a sudden I'm in jhb ?! o_O. 160ms ping
Here is the link : http://www.speedtest.net/result/985318087.png

Also only got 0.40mb\s download and 0.41 upload o_O ?!
 
Last edited:
@Gordon_R - The CBD seems fine - I just RDP'ed to my server at work and I get a 28 ping to an IS gaming server from there - the same IP gives me a ping of 220 right now from my home in Newlands.

btw my work is in De Waterkant
 
I'm just wondering, but can telkom automatically downgrade your line speed? Reason i ask is i previously synced at 10mb when it came out and looking now im sitting at 3890, thinking that maybe telkom messed something up over there.
 
@acidrain If the sync isn't reliable at the configured line speed it will auto negoiate down in increments.
 
Im in Retreat, Cape Town on a 4 meg, isp Mweb

Sun Oct 10 2010 20:57:55 GMT+0200 (South Africa Standard Time)
PING mweb.co.za (196.2.63.110) 56(84) bytes of data.
64 bytes from www.mweb.co.za (196.2.63.110): icmp_seq=1 ttl=123 time=9.99 ms
64 bytes from www.mweb.co.za (196.2.63.110): icmp_seq=2 ttl=123 time=10.3 ms
64 bytes from www.mweb.co.za (196.2.63.110): icmp_seq=3 ttl=123 time=10.9 ms
64 bytes from www.mweb.co.za (196.2.63.110): icmp_seq=4 ttl=123 time=10.2 ms
64 bytes from www.mweb.co.za (196.2.63.110): icmp_seq=5 ttl=123 time=9.82 ms

--- mweb.co.za ping statistics ---
5 packets transmitted, 5 received, 0% packet loss, time 4003ms
rtt min/avg/max/mdev = 9.825/10.282/10.943/0.390 ms

but last week i couldn't even run my vpn, so it will be interesting to see what happens tomorrow morning.
 
I'm not sure whether to be proud or jealous :) That's great feedback proguy. If you get a moment could you show us some tracerts to one or two local destinations.

It would be very useful for the MWEB users in the problematic areas to see what it should look like.
 
Last edited:
Sun Oct 10 2010 21:17:45 GMT+0200 (South Africa Standard Time)
traceroute to mweb.co.za (196.2.63.110), 30 hops max, 40 byte packets
1 41-133-108-1.dsl.mweb.co.za (41.133.108.1) 8.705 ms 10.668 ms 12.638 ms
2 196-28-178-202.adsl.mweb.co.za (196.28.178.202) 16.610 ms 18.584 ms 21.554 ms
3 196.41.144.36 (196.41.144.36) 24.526 ms 26.498 ms 29.472 ms
 
@Will. A bit off-topic, but I noticed that MWeb's IPC now routes via CINX (CT), and posted this on the 'big' thread. Potentially this would allow faster connection to Cape Town sites.

Based on proguy's post I checked and found that www.mweb.co.za is only 5 hops away. Is the MWeb site also being mirrored in Cape Town now? (The DNS and SMTP mail servers are still based in Joburg).
 
SMTP server

Sun Oct 10 2010 21:43:42 GMT+0200 (South Africa Standard Time)
PING smtp.mweb.co.za (196.28.80.20) 56(84) bytes of data.
64 bytes from relay.smp.mweb.co.za (196.28.80.20): icmp_seq=1 ttl=122 time=33.9 ms
64 bytes from relay.smp.mweb.co.za (196.28.80.20): icmp_seq=2 ttl=122 time=34.5 ms
64 bytes from relay.smp.mweb.co.za (196.28.80.20): icmp_seq=3 ttl=122 time=33.9 ms
64 bytes from relay.smp.mweb.co.za (196.28.80.20): icmp_seq=4 ttl=122 time=33.3 ms
64 bytes from relay.smp.mweb.co.za (196.28.80.20): icmp_seq=5 ttl=122 time=35.8 ms

--- smtp.mweb.co.za ping statistics ---
5 packets transmitted, 5 received, 0% packet loss, time 4003ms
rtt min/avg/max/mdev = 33.393/34.321/35.832/0.859 ms
 
@Gordon

Yes we're now carrying a significant amount of traffic on CINX, as we are on JINX up north which helps a lot, as you pointed out, in terms of reducing hops and latency to local content. I'm speaking under correction here (things move a little fast sometimes for me even to keep up :) ), but I think what you're seeing there is the fact that the portal servers are in fact in the Cape Town data center. In terms of the DNS and SMTP platforms I know there has been some discussion around the possibility of mirroring that in Cape Town, but I think that project is still in the conceptual phase.
 
Sunnydale,fish hoek,Cape Town on a 4 meg, isp Mweb uncapped (021785****)

Tracing route to 41.185.60.18 over a maximum of 30 hops

1 <1 ms <1 ms <1 ms home.gateway [10.0.0.2]
2 * * * Request timed out.
3 9 ms 9 ms 8 ms 196-28-178-166.adsl.mweb.co.za [196.28.178.166]

4 9 ms 9 ms 9 ms 196-28-178-133.adsl.mweb.co.za [196.28.178.133]

5 11 ms 10 ms 10 ms 198.32.214.45
6 12 ms 11 ms 10 ms vl40.cr.gw.cpt.za.wa.co.za [41.185.0.97]
7 12 ms 12 ms 11 ms 41.185.60.18

Trace complete.


Tracing route to www.mweb.co.za [196.2.63.110]
over a maximum of 30 hops:

1 <1 ms <1 ms <1 ms home.gateway [10.0.0.2]
2 * * * Request timed out.
3 8 ms 11 ms 8 ms 196-28-178-202.adsl.mweb.co.za [196.28.178.202]

4 9 ms 9 ms 9 ms 196.41.144.36
5 10 ms 9 ms 11 ms www.mweb.co.za [196.2.63.110]

Trace complete.

looks good to me.
 
Last edited:
Top
Sign up to the MyBroadband newsletter
X