NTL 10mbit connection is appalling

[dunkyb]

Hi,

I have had NTL cable broadband for quite a long time now. Since upgrading to 10mbit, if uploading at even 20Kbyte/s (out of the possible ~60) I see packet loss of up to 50% to my gateway (First hop, NTL side).

Even without uploading a thing (I don't do it, my brothers are to blame!) latency on the connection is awful. I have never seen a connection fluctuate as much as mine... Say, between 9ms and 50ms to yahoo.co.uk ? It is highly unstable.

I have had engineers out, at one point I even had a complete cable re-pull (which he made a hash of, wish I hadn't bothered). I had an NTL: 250 cable modem installed as the old 120 couldn't cope with the 10mbit. I have phoned CS/TS numerous times and they say the connection is fine.

My largest concern is web browsing, is awful. A lot of the time it hangs/stalls, yet a simple refresh will load the page.

Please advise ... I am in the Portsmouth area, and go through the Cosham POP.

Kind regards

Duncan B.

[Chrysalis]

try ntl responsetek, your problem I think is getting them to acknowledge a problem which may well be ubr saturation. I think I remember ignition saying packet loss at the ubr means the cpu is maxed out, but you should wait for someone who has/does work for ntl to respond for a better diagnosis.

[dcclanuk]

guys my 10mb has never reached full potential, its *seems* to be capped at around 5-6mb... i remember there is a site to chk with some user name and pw at wat speed u are synced that..........wat is that site and username/pw again?

also, wat else could be the prob? could it be the fact that i am using the USB instead of the ETHERNET???

my area is LWWEM << that was the bt exchange anyway!

[SnoopZ]

http://192.168.100.1/ username root ,password root

Always best to use ethernet.

[dunkyb]

Originally Posted by Chrysalis

try ntl responsetek, your problem I think is getting them to acknowledge a problem which may well be ubr saturation. I think I remember ignition saying packet loss at the ubr means the cpu is maxed out, but you should wait for someone who has/does work for ntl to respond for a better diagnosis.

Hi,

Yeah I'm fairly sure it is just saturation on the upstream, but every time I've phoned up they've said its fine .... I think NTL's idea of fine is perhaps different to my own, however.

What is NTL responsetek, sorry?

Someone from NTL tech support tried telling me that the UBRs dont prioritise ICMP traffic, so ping results are not a good indication of the connection/latency. This would all be very well, if UNIX traceroute didn't use UDP! Plus, I think if a router is struggling to respond to ICMP whatever QoS level its given, it's probably being overloaded?

If someone from NTL could have a look at my UBR (ubr01horn-ge02.inet.ntl.com) and give an honest opinion on the upstream usage, I'd be really greatful.

Cheers

[dcclanuk]

Originally Posted by SnoopZ

http://192.168.100.1/ username root ,password root

Always best to use ethernet.

thx for that.. chked it out.. its 10mb/512k , but i am sure ethernet wont make a difference of 3-4 mb extra???

could it be my congested UBR? i think it is something like brent.something like that

[Chrysalis]

well the old dont prioritise icmp is a common excuse but funnily enough icmp only seems to jump when the router is too busy and when its too busy is that good?

I think it might be cpu satuartion rather then upstream saturation as ntl seem to know about upstream problems more rather then cpu and downstream problems, have you tried cycling your modem to get it on a new upstream channel?

[Graham M]

USB1 (which the modem will be) only supports 12MBPS across all ports on the hub, so with a couple of devices on the hub, and then factor in overheads, 4-6MBit difference is certainly plausable

[dunkyb]

Originally Posted by Chrysalis

well the old dont prioritise icmp is a common excuse but funnily enough icmp only seems to jump when the router is too busy and when its too busy is that good?

I think it might be cpu satuartion rather then upstream saturation as ntl seem to know about upstream problems more rather then cpu and downstream problems, have you tried cycling your modem to get it on a new upstream channel?

Yeah the connection is usually fine when pings are steady... Which unfortunately isn't too often.

It's 1:30am (I'm at work on nights, not sad!) and I was just seeing 5% packet loss to my gateway! The connection is just sat idle.

I've tried power cycling the modem loads of times, even leaving it off over night.. it always seems to home back in on the same upstream (6, at the moment.. U/S power level 46dB)

[James Henry]

Probably not relevant but whether a traceroute uses UDP or ICMP both are on a low priority to routers, indeed any non routing system related traffic directed towards the router's own address, especially when they are working on other CPU intensive things.

[monkey2468]

Originally Posted by dunkyb

Yeah the connection is usually fine when pings are steady... Which unfortunately isn't too often.

It's 1:30am (I'm at work on nights, not sad!) and I was just seeing 5% packet loss to my gateway! The connection is just sat idle.

I've tried power cycling the modem loads of times, even leaving it off over night.. it always seems to home back in on the same upstream (6, at the moment.. U/S power level 46dB)

I have had a look at your UBR and downsteam and upstream look fine. Max CPU utilisation has not been over 64% and average is 42%. If you PM me your full postcode and house number I can look into a bit more for you.

[dunkyb]

Thanks, PM'd.

---------- Post added at 18:47 ---------- Previous post was at 18:26 ----------

Noticed my U/S power level is some what higher than usual tonight, ~50dBmv as opposed to 46.. also D/S is -5, usually -4.7 ish ....

Not sure if this is in light of the packet loss probs or not.

P.S. Took 3 attempts to post this