Network problems Tonight ? - Page 11

IanUK · 29-08-2005, 20:10

Even worse now

Target Name: www.froopter.com
IP: 66.98.246.48
Date/Time: 29/08/2005 21:03:07
2 7 ms [10.82.176.1]
3 6 ms wapk-t2cam1-a-ge-wan41-112.inet.ntl.com [80.1.170.57]
4 11 ms popl-t2core-a-ge-wan64.inet.ntl.com [62.255.81.65]
5 7 ms pop-bb-a-so-330-0.inet.ntl.com [213.105.174.113]
6 10 ms pop-bb-b-ae0-0.inet.ntl.com [213.105.172.77]
7 11 ms tele-ic-1-so-010-0.inet.ntl.com [62.253.185.82]
8 210 ms [212.250.14.62]
9 275 ms p10-0.core01.bos01.atlas.cogentco.com [130.117.0.46]
10 303 ms p5-0.core01.ord01.atlas.cogentco.com [66.28.4.110]
11 295 ms p15-0.core02.ord01.atlas.cogentco.com [66.28.4.62]
12 306 ms p12-0.core01.mci01.atlas.cogentco.com [66.28.4.33]
13 322 ms p12-0.core01.iah01.atlas.cogentco.com [66.28.4.105]
14 322 ms g0-2.na21.b015619-0.iah01.atlas.cogentco.com [66.28.64.66]
15 327 ms everyones_internet_2.demarc.cogentco.com [38.112.25.158]
16 133 ms gphou-66-98-241-27.ev1.net [66.98.241.27]
17 132 ms gphou-66-98-241-125.ev1.net [66.98.241.125]
18 139 ms srv1.froopter.com [66.98.246.48]

And my test file from my other Telia USA hosted server is less than 15k now, pathetic again.

As usual - direct connections, or AOL proxies bring both still in at full speed, I think its time someone from NTL did something about this

Chrysalis · 29-08-2005, 20:10

Quote:

Originally Posted by Ignition

Just to be fair this isn't my fix I don't work for ntl anymore

umm that puts me up **** creek

IanUK · 29-08-2005, 20:19

The Cogent issue is separate to the proxy issue I think, it only affects my server through the web browser because the Poplar Proxy chooses that route, direct connections go through Abovenet and don't suffer the same.

My other server is different though, traces look ok, but speeds through NTL proxies are terrible, All other ISP's seem fine, just NTL users that are having problems connecting to my server(s).

Direct connections show the bandwidth is there, so its not congestion this end, it looks like NTL's peering is pretty poor, this is two networks now that are suffering, its either that or the proxies, I don't know

I ask again, why don't NTL monitor this kind of thing ?
The Cogent fix is coming we are told, and it was fixed temporarily, so why can't it be fixed temporarily again, and keep fixing it temporarily until the fibre arrives ?

This has been going on for weeks now.
I see yet another proxy problem thread started today as well (http://www.cableforum.co.uk/board/sh...ad.php?t=34293), thats at least one a day recently, NTL fix these things PLEASE !!!!!

Chrysalis · 29-08-2005, 20:34

The cogent issue is defenitly nothing to do with the proxies, I noticed it when there was a massive pause waiting for my ssh session to respond, ssh traffic is direct and doesnt use proxies. I havent done any speed tests over ftp. This kind of thing I agree with you I expect to be monitored so I am guessing one of 2 things, ntl have waited until its full to the brim before the new fiber got laid, or these recent high speeds people have been reporting caused a surge in the utilisation.

Stuart · 29-08-2005, 20:39

Just to back up IanUK, I have no proxy set, yet when I use VisualRoute to trace the route to that site, the ping time rises at the last point in the route that's in the NTL network. According to NTL, this point is where NTL peer to Cogent.

jonholyfield · 29-08-2005, 21:14

Quote:

Originally Posted by Chrysalis

The cogent issue is defenitly nothing to do with the proxies, I noticed it when there was a massive pause waiting for my ssh session to respond, ssh traffic is direct and doesnt use proxies. I havent done any speed tests over ftp. This kind of thing I agree with you I expect to be monitored so I am guessing one of 2 things, ntl have waited until its full to the brim before the new fiber got laid, or these recent high speeds people have been reporting caused a surge in the utilisation.

I agree.

This stinks of bad management.

The market is horribly competitive.

I dont know but it seems as though they have looked at the competition and are being told to deliver the same. Bad move as far as I can see. Pipex are giving me 2 meg with no problem but they never promised it.

Interesting distiction there when they agree a one meg connection and you get 2 you think "Oh Thanks". Promise ten and degrade the whole network and nobody is happy.

Chrysalis · 29-08-2005, 21:52

Quote:

Originally Posted by jonholyfield

I agree.

This stinks of bad management.

The market is horribly competitive.

I dont know but it seems as though they have looked at the competition and are being told to deliver the same. Bad move as far as I can see. Pipex are giving me 2 meg with no problem but they never promised it.

Interesting distiction there when they agree a one meg connection and you get 2 you think "Oh Thanks". Promise ten and degrade the whole network and nobody is happy.

Well whats baffling here and with IanUK's EV1 case I was originally wrong and NTL decide on their outbound route. EV1 and staminus both have different providers, abovenet is very good to staminus, was less hops and much lower latency. nrs which kicked in the last 4-6 weeks was lower standard but still only 40ms over abovenet and useable, so ntl not only are using a congested link but are also preferencing it over non congested routes.

IanUK · 30-08-2005, 07:25

I'm guessing its cost, Cogent is well known for being a cheap provider, most 'pile em high' type internet businesses (Newshosting/Newsreader/Filefront/EV1) will have Cogent bandwidth for their High Usage situations and I'm guessing its also cheaper for NTL to route using Cogent.

None of this helps us poor customers of course

And other ISP's seem to manage with Cogent just fine.

Perhaps Telia is a cheap route as well ? might explain why it also suffers from about 12pm onwards....

In fact after more research it seems that NTL's peering is pretty poor all round, Easynews has a route selector tool that lets you choose the network that carries your bandwidth, their support forum is full of uk cable users that have very poor connections to them using virtually all of the networks, Level 3 works great for me to them, full speed, but the other providers (Qwest,LLNW,UU & ATT) are all very poor (mid 100k's) so thats another 4 providers that NTL seem to have poor speeds with.

I think its looking to be network probs and not proxies...it just shows as the proxies because different proxies can take different routes...or maybe it *is* the proxies, I don't know, we can witter on for ages here talking to ourselves but we need someone from NTL to answer my Responstek as promised !!

jtwn · 30-08-2005, 10:35

I just realised that my old news server takes the same route, its mighty strange that i used to go through the same 'mental' hop as you but could flat line 360kb/s with four threads.

That was from February for some months...something happened on ~16th June, could not get over 130kb/s even with 8 threads. They upgraded something, unrelated really and soon after I could max out with 8 threads, on occasions it was flaky between 320 and 350 but nothing to shout about.

Chrysalis · 30-08-2005, 16:46

Quote:

Originally Posted by IanUK

I'm guessing its cost, Cogent is well known for being a cheap provider, most 'pile em high' type internet businesses (Newshosting/Newsreader/Filefront/EV1) will have Cogent bandwidth for their High Usage situations and I'm guessing its also cheaper for NTL to route using Cogent.

None of this helps us poor customers of course

And other ISP's seem to manage with Cogent just fine.

Perhaps Telia is a cheap route as well ? might explain why it also suffers from about 12pm onwards....

In fact after more research it seems that NTL's peering is pretty poor all round, Easynews has a route selector tool that lets you choose the network that carries your bandwidth, their support forum is full of uk cable users that have very poor connections to them using virtually all of the networks, Level 3 works great for me to them, full speed, but the other providers (Qwest,LLNW,UU & ATT) are all very poor (mid 100k's) so thats another 4 providers that NTL seem to have poor speeds with.

I think its looking to be network probs and not proxies...it just shows as the proxies because different proxies can take different routes...or maybe it *is* the proxies, I don't know, we can witter on for ages here talking to ourselves but we need someone from NTL to answer my Responstek as promised !!

The proxies have issues, they just different issues.

right now the cogent problem is back to normal, but I wonder what it will be like tonight.

So much of usa is turning to cogent now which is very scary since the provider is still below the standards of AT&T abovenet etc. However many providers still have it amongst a mix of providers so that does still leave the likes of NTL with a choice of route, so I am very dissapointed NTL have opted for budget over quality since they sell the "we cap you but it improves quality" so much.

Chrysalis · 30-08-2005, 20:46

its back as predicted.

6 18 ms 17 ms 20 ms tele-ic-1-so-010-0.inet.ntl.com [62.253.185.82]

7 243 ms 221 ms 216 ms 212.250.14.62
8 285 ms 294 ms 290 ms p10-0.core01.bos01.atlas.cogentco.com [130.117.0
.46]
__________________

ok staminus have had me run some tests and have done something their end to help force change the route, it hasnt kicked in yet and wont kick in if ntl have cogent forced.

test #1 to abovenet ntl to above handoff - 60ms - poor - possible congestion

5 15 ms 14 ms 17 ms pop-bb-b-so-100-0.inet.ntl.com [62.253.185.238]

6 63 ms 76 ms 63 ms so-4-0-0.cr2.lhr3.uk.above.net [208.185.188.41]

7 167 ms 111 ms 112 ms so-7-0-0.cr2.lga1.us.above.net [64.125.31.182]

test #2 ntl to nrs handoff - 90ms - very poor likely congestion
8 19 ms 19 ms 17 ms tele-ic-2-so-110-0.inet.ntl.com [62.253.185.74]

9 86 ms 99 ms 123 ms cr02.ldn01.pccwbtn.net [195.66.224.167]

test #3 ntl to cogent - very poor - 200ms - confirmed congestion

6 16 ms 20 ms 15 ms tele-ic-1-so-010-0.inet.ntl.com [62.253.185.82]

7 195 ms 201 ms 201 ms 212.250.14.62
8 289 ms 284 ms 286 ms p10-0.core01.bos01.atlas.cogentco.com [130.117.
.46]

looks like with all the ubr upgrades going on ntl forgot about their peering, no wonder they keeping the proxies enabled, external traffic is pretty clogged.

Chrysalis · 30-08-2005, 23:03

problems easing a bit.

6 17 ms 15 ms 23 ms tele-ic-1-so-010-0.inet.ntl.com [62.253.185.82]

7 162 ms 164 ms 169 ms 212.250.14.62
8 217 ms 221 ms 216 ms p10-0.core01.bos01.atlas.cogentco.com [130.117.0

abovenet looks clear now

5 15 ms 14 ms 16 ms pop-bb-b-so-100-0.inet.ntl.com [62.253.185.238]

6 20 ms 21 ms 66 ms so-4-0-0.cr2.lhr3.uk.above.net [208.185.188.41]

Chrysalis · 31-08-2005, 00:02

at 1am cogent looks normal.

6 34 ms 19 ms 25 ms tele-ic-1-so-010-0.inet.ntl.com [62.253.185.82]

7 22 ms 18 ms 59 ms 212.250.14.62
8 90 ms 90 ms 91 ms p10-0.core01.bos01.atlas.cogentco.com [130.117.0
.46]

IanUK · 31-08-2005, 07:40

Looks like its Cogent/Telia/Abovenet & PCCWBTN thats proven to be getting congested then

Just to keep everyone up to date - I've had two replies to my Responsetek form, and am now awaiting a call from the broadband department.

I'm not sure why they have to call me when all the info is in this thread and affects other people than me, but thats fine if thats how they want to do it.

I've been told I'll be contacted by this Friday, I'll post here when I get more details.

**Paul** · 31-08-2005, 07:46

Quote:

Originally Posted by Stuart C

Just to back up IanUK, I have no proxy set, yet when I use VisualRoute to trace the route to that site, the ping time rises at the last point in the route that's in the NTL network. According to NTL, this point is where NTL peer to Cogent.

No surprise, since at that point you are crossing the pond, so the ping times will rise by quite a bit.