Is UUNET the problem?

[leewsimpson]

I am investigating slow internet speeds from IBurst. Please help me understand my findings correctly, and please point me in the right direction as to how I can help resolve these issues.

Here is a typical trace from my machine to a sample USA ip address:

Target Name: www.ea.com
IP: 159.153.253.82
Date/Time: 2005/06/12 01:48:51 PM to 2005/06/12 01:49:10 PM

Hop Sent Err PL% Min Max Avg Host Name / [IP]
1 20 0 0.0 41 813 271 [196.30.31.100]
2 20 0 0.0 55 1181 315 [192.168.0.97]
3 20 0 0.0 49 744 279 ge0-2.gw10.jnb6.alter.net [66.8.42.21] -- UUNET-InternalNET
4 20 0 0.0 47 1113 334 fe4-1-0.cr1.jnb6.alter.net [196.30.156.33] -- UUNET-InternalNET
5 20 0 0.0 287 1312 487 atm1-1-0sub902.ir2.mia16.alter.net [196.30.229.209] -- UUNET-InternalNET
6 20 1 5.0 271 1275 464 POS0-1-0.IH4.MIA4.ALTER.NET [152.63.86.145] -- USA
7 20 2 10.0 260 1235 468 202.at-5-1-0.CL2.MIA4.ALTER.NET [152.63.7.130]
8 20 1 5.0 262 1210 444 0.so-4-2-0.XL2.MIA4.ALTER.NET [152.63.101.46]
9 20 0 0.0 288 1175 460 0.so-4-2-0.XL2.ATL5.ALTER.NET [152.63.81.81]
10 20 0 0.0 284 1140 451 0.so-7-0-0.BR3.ATL5.ALTER.NET [152.63.87.93]
11 20 1 5.0 385 1124 572 sl-bb20-atl-11-2.sprintlink.net [144.232.18.13]
12 20 1 5.0 404 1083 555 sl-bb21-atl-14-0.sprintlink.net [144.232.12.142]
13 20 1 5.0 396 1044 549 sl-bb21-rly-8-0.sprintlink.net [144.232.20.177]
14 20 0 0.0 466 1236 638 sl-bb23-rly-11-0.sprintlink.net [144.232.14.134]
15 20 0 0.0 418 1035 561 sl-st20-ash-11-0.sprintlink.net [144.232.20.150]
16 20 0 0.0 385 993 552 sl-nufx-11-0.sprintlink.net [144.223.246.6]
17 20 0 0.0 402 957 577 fert02.abn-iad.ea.com [159.153.224.11]
18 20 0 0.0 440 1067 626 [159.153.224.34]
19 20 0 0.0 410 1043 637 [159.153.253.7]
20 20 0 0.0 426 1467 639 napdhome.pogo.com [159.153.253.82]


The main area of concern is of course the UUNET internal net IP pings. They are consistently higher, and that’s even before it gets outside of South Africa. I cannot seem to find out where the "InternalNet" actually resides, although I suspect this could be Cape Town.

Resources:
www.pingplotter.com
http://www.uunet.co.za/_about/network/diagram/default.asp?n=953
http://www.rwhois.net/rwhois/prwhois.html
http://www.afrinic.net/cgi-bin/whois

 

[slimothy]

in short no, you see if it were UUNET then the slow speeds would affect everyone, but they don't, after months of reading peoples problems here I can say that 99% of speeds problems are down to 2 things, people may suffer from one, both or non of these things.

1) bad quality of signal, either bad signal or high frame error rate
2) tower congestion, if your tower is packed to the max you will get crap speeds while people are all fighting over the bandwith. This is the main problem, this is why my tower in Kempton park is nice and fast and Northcliff for example is slow especially in peak hours like the day time. Alot of users have also noticed speed increases on towers like Northcliff in the early hours of the morning so it makes sense the problem is congestion.

You mentioned you were investigating the reasons for slow speed, well pings won't help you there, latency isn't really a factor, even if you are on a service with bad latency like satelite internet you can still get fast speeds. If it takes 4 seconds for example for data to travel from point a to point b but the pipe is 10mbit in diameter then you would still get 10mbit speeds even though the latency is crap.

If you get slow speeds and you have full signal and low frame error rate then it is more likely reason 2 I mentioned above so the only way to solve that problem is to get WBS to install another tower in that area (not gonna happen) or changing the tower you connect to by either moving or getting the external antenna and pointing to a diff tower.

 

[Ravilj]

I can confirm what slimothy said as well, my speeds pick up nicely from bout 10pm onwards during the day well lets not go there. I connect to the Bryanston tower.

 

[Clipse]

to be honest, I think iburst is using the same link uunet uncapped is on atm.. higher ping aswell as more packetloss, but http traffic works well and local latency is consistent tho. Can you iburst users confirm on international traceroutes that you are getting packetloss , or more packetloss than before?

 

[slimothy]

thing is WBS use IS and UUNET and the problems do seem to be tower specific so if that was a problem, the UUNET uncapped link, it would only be for certain users at certain times so how could anyone confirm that. Do people suffer from packet loss on iBurst? Hell yes, and more than just packet loss but like I said its only certain people at certain times. Theres so many ingredients in the pot that its hard to pick which one smells like ass

 

[guest2013-1]

WBS does not use Internet Solutions. Don't know where people got that in their head

 

[ic]

It was forumites saying the Helpdesk told them that WBS were going to be smoking with a pipe to IS, there was never any real evidence that it actually happened - this goes back to problems around launch-time ~April...

 

[slimothy]

No, that extra 15MB pipe they installed in April was from IS

 

[ic]

No, that extra 15MB pipe they installed in April was from IS

You don't find it odd that tracert's to sites on the IS backbone never show any trace of a direct route to IS that doesn't first go through UUNET...?

 

[slimothy]

i dont know what they used the pipe for, but i know they got the pipe, for all you or i know they use it for thier own internal network or cape town users

 

[ic]

i dont know what they used the pipe for, but i know they got the pipe, for all you or i know they use it for thier own internal network or cape town users

I think they use it as a doob-bomb & most of the times when I don't experience problems with iBurst it's bcos the WBS techies are too stoned to mess with stuff that was working...

 

[seburn]

Large memory cap for a pipe bomb - must be quite a firework show - all those instructions on how to blow up!

 

[Crash]

The company I work for have Servers hosted at UUNET. When I have a major problem on iBurst, I connect to the servers and check from there. I haven't had an occasion were there were internet connection problems from the Servers. So I also think it's contention on towers and Signal problems.

 

[Kei]

1) Signal problems due to weather conditions and background noise. EN norms are not enforced in SA so I am sure there's lots of electronic stuff in use in the field generating EMI

2) Peering links between towers are RF/microwave. Signal gets degraded due to atmospheric effects

3) Links between towers and WBS offices having problems as I suspect they are Telkom diginet lines, the usual problems most Diginet line users experience i.e. pair theft, running errors, wet cables, et cetera...

4) Problems inside WBS's network. bandwidth manager/shaper causing problems. Network machines inadequate or too small.

5) Pipe to IS itself

 

[slimothy]

1) thats just signal issues, and I can't say weather has ever caused me any problems, I was under the impression that it had very limited effects and that rain etc didn't affect it at all.

2) thats basically the same at #1

3) WBS is Wireless Business Solutions, they have a licence for wireless networking in this country and already run the wireless lotto networks so I would assume they would use wireless solutions as much as possible on thier network because it would be cheaper and they already have expertise in that field. I think the link between base stations and thier offices is entirly thier own infrustructure. If it were telkom diginets, i'm not sayig i think it is, but if it was, all speed on a diginet is GUARANTEED, which means if its a 100mbit link, that 100mbit is guaranteed, anytime.

4) this would mean the problems would affect everyone since everyone's traffic goes through the same gateway and thats where the intelligent firewall is implemented.

5) the primary pipes are UUNET, as far as i know the IS pipe was used to supply a bit of extra bandwith and its only a 15MB pipe so even if it closed completly the network could still operate 100%.

 

[seburn]

1. For ppl in other areas or on borders signal is still an issue

2. for some ppl who connect to tower that are peering (some aren't) I think the peering further cloggs another tower say northclif -atm eff ect are End user problem WBS will have good links in place.

3. Not a problem

4. aggree with slim

5. aggree with slim

so what we are saying is 80% of problem is contention 20% is End user signal which btw contention effects adversely - which means 99.999999% is contention