Looks like VM taken congestion to a new level - leicester west

[Chrysalis]

Hi guys I am back.

Ok so I have had a new cable connection now since january and initially was good but has noticebly and steadily got worse, I havent immediatly noticed the severeity of it until yesterday tho due to been using my adsl for all my normal use (previously just tested the cable at random times).

To cut short the connection is not good enough to warrant cancelling my adsl.

Yesterday at 10am-11am I had pings that were hovering between 10 and 200ms to bbc.co.uk and eventually they were pegged above 100ms, not a single ping below 100ms.

Speedtests were hovering between 1 and 3mbit/sec (on 20mbit package) and when the pings were pegged at 100ms+ my speedtest was 400KBIT/sec 0.4meg. less than my upload. 10am is a fairly quiet time of date, very worrying. My modem was also dropping out tho and someone has advised me my power levels are too high so this will be reported tommorow after 9am and I will get it fixed.

However there is signs this is at least partially congestion related, tonight I stayed up and have been running tests.

After 10pm my speeds were higher than the 10am speeds around 4-7mbit, pings fluctuating all over the place but better than the daytime.
After midnight it is steadily and sharply improving, pings almost normal but still some fluctuation (even when most people are asleep oh dear) and I can get speeds of around 15mbit. Unless my power levels have improved during the night it looks like VM(ntl) overselling like crazy again.

Some questions if people dont mind answering.

1 - Will changing to a different tier have a chance of fixing this? I am aware of docsis3, it is used on 50mbit, I assume going to 50mbit garnatuees I am on docsis3 or some areas dont have docsis3 yet? will changing to 10mbit also put me on docsis3? have read in few places funny enough 10mbit gets docsis3 but 20mbit doesnt, would like this confirmed if true or not. My sister said she had no congestion on 10mbit, then she upgraded and has the same problems as me.
2 - Why do VM let some areas rot whilst others have great service? people have been showing me pings that are fine any time of the day and say their speeds are great 24/7, yet my area (and my old area) seem to be only able to max out about 2-3 hours in the early morning. At least my old area wasnt as slow as this one tho 400kbit out of 20mbit ouch.
3 - given that I have seen pings and speed improve after midnight is there still a chance this is all caused by my powerstream level?

10AM speed and pings

0.4mbit down
0.6mbit up

Pinging bbc.co.uk [212.58.224.138] with 32 bytes of data:
Reply from 212.58.224.138: bytes=32 time=132ms TTL=120
Reply from 212.58.224.138: bytes=32 time=94ms TTL=120
Reply from 212.58.224.138: bytes=32 time=106ms TTL=120
Reply from 212.58.224.138: bytes=32 time=108ms TTL=120
Reply from 212.58.224.138: bytes=32 time=103ms TTL=120
Reply from 212.58.224.138: bytes=32 time=110ms TTL=120
Reply from 212.58.224.138: bytes=32 time=111ms TTL=120
Reply from 212.58.224.138: bytes=32 time=88ms TTL=120
Request timed out.
Reply from 212.58.224.138: bytes=32 time=89ms TTL=120
Reply from 212.58.224.138: bytes=32 time=40ms TTL=120
Reply from 212.58.224.138: bytes=32 time=88ms TTL=120
Reply from 212.58.224.138: bytes=32 time=77ms TTL=120
Reply from 212.58.224.138: bytes=32 time=129ms TTL=120

1am speed and pings

15mbit down
0.6mbit up

Pinging bbc.co.uk [212.58.224.138] with 32 bytes of data:
Reply from 212.58.224.138: bytes=32 time=10ms TTL=118
Reply from 212.58.224.138: bytes=32 time=9ms TTL=118
Reply from 212.58.224.138: bytes=32 time=22ms TTL=118
Reply from 212.58.224.138: bytes=32 time=11ms TTL=118
Reply from 212.58.224.138: bytes=32 time=10ms TTL=118
Reply from 212.58.224.138: bytes=32 time=14ms TTL=118
Reply from 212.58.224.138: bytes=32 time=9ms TTL=118
Reply from 212.58.224.138: bytes=32 time=11ms TTL=118
Reply from 212.58.224.138: bytes=32 time=16ms TTL=118
Reply from 212.58.224.138: bytes=32 time=11ms TTL=118
Reply from 212.58.224.138: bytes=32 time=23ms TTL=118
Reply from 212.58.224.138: bytes=32 time=10ms TTL=118
Reply from 212.58.224.138: bytes=32 time=12ms TTL=118
Reply from 212.58.224.138: bytes=32 time=10ms TTL=118
Reply from 212.58.224.138: bytes=32 time=11ms TTL=118

[marky]

I like this test (http://www.pingtest.net/) it was a great help over the last few days trying to get VM to sort mine out (10mbps)
Not the best but its a A :D
http://www.pingtest.net/result/13743318.png (http://www.pingtest.net)

[Chrysalis]

yeah I will run that today when things congest again.

although I have never heard of VM acknowledging jitter as a problem.

[marky]

yeah I will run that today when things congest again.

although I have never heard of VM acknowledging jitter as a problem.
Tech support gave me the speedtest one over the phone, I found that on the same page and just quoted stuff back.

---------- Post added at 07:55 ---------- Previous post was at 07:51 ----------

Like this one.
http://www.pingtest.net/result/13644903.png

---------- Post added at 07:58 ---------- Previous post was at 07:55 ----------

I think it was the packet loss that sold it to them.

[Chrysalis]

ok 8am and already signs of congestion.

https://www.cableforum.co.uk/images/local/2010/03/5.png (http://www.speedtest.net) -after 2am
https://www.cableforum.co.uk/images/local/2010/03/6.png (http://www.speedtest.net) -8am

jitter according to pingtest still not too bad.

http://www.pingtest.net/result/13750380.png (http://www.pingtest.net)

although my pings to bbc.co.uk are not pretty at the moment.

Pinging bbc.co.uk [212.58.224.138] with 32 bytes of data:
Reply from 212.58.224.138: bytes=32 time=27ms TTL=118
Reply from 212.58.224.138: bytes=32 time=21ms TTL=118
Reply from 212.58.224.138: bytes=32 time=20ms TTL=118
Reply from 212.58.224.138: bytes=32 time=26ms TTL=118
Reply from 212.58.224.138: bytes=32 time=28ms TTL=118
Reply from 212.58.224.138: bytes=32 time=16ms TTL=118
Reply from 212.58.224.138: bytes=32 time=19ms TTL=118
Reply from 212.58.224.138: bytes=32 time=16ms TTL=118
Reply from 212.58.224.138: bytes=32 time=22ms TTL=118
Reply from 212.58.224.138: bytes=32 time=14ms TTL=118
Reply from 212.58.224.138: bytes=32 time=31ms TTL=118
Reply from 212.58.224.138: bytes=32 time=34ms TTL=118
Reply from 212.58.224.138: bytes=32 time=20ms TTL=118
Reply from 212.58.224.138: bytes=32 time=28ms TTL=118
Reply from 212.58.224.138: bytes=32 time=12ms TTL=118
Reply from 212.58.224.138: bytes=32 time=24ms TTL=118
Reply from 212.58.224.138: bytes=32 time=19ms TTL=118
Reply from 212.58.224.138: bytes=32 time=25ms TTL=118

[Ignitionnet]

See my Think Broadband posts :p:

You may also want to post on http://community.virginmedia.com - the newsgroups replacement.

[Chrysalis]

the tech is coming tommorow morning iginition booked now :) , I am really hoping this is all down to the power level as thats an easy fix. For whats it worth my modem is also making a quiet buzzing noise.

at 8am it didnt look good but today it seems not so bad yet as it hasnt got worse since 8am to now. nowhere near as bad as yesterday at the moment.

already posted on that community site as well :)

[lyniroquai]

I had a problem with my download speed too, I rang Virgin media today 8 times to try sort it out and they just kept telling me they would send a new signal out or that I would have to sit hardwired to my laptop constantly just to get my browser to load!

I finally got through to complaints and they redirected me to technical support, I was only getting a download speed of 0.07 !!!

They told me to hardwire my laptop and then do another speed test and the download went up to 19.XX.

The technician told me it was because of interference in my area and that I should change my channel on my wireless router settings so I tried every channel and checked with speed test and I finally found the right channel which works perfectly now

[Chrysalis]

this isnt anything local.

the modem itself is hard wired connected to my equipment. the problem for VM is they will have a hard time blaming my pc because I have a 2nd internet connection in my property.

so the tech coming today I will ask him questions about how many customers are on the ubr etc. if he even hints at blaming my pc I will just plug in my adsl easynet connection and do a demonstration.

last night it was aweful, I wasnt on my pc during the evening and I came on at about 1.30am. Typically most people are offline and asleep then. speeds around 5-7mbit, pings all over the place and next to no pings normal, I have it all graphed on thinkbroadband's quality graph service. I hope this is down to the powerlevel because if it is congestion I dread to think how much they have oversubscribed the area for it to be congested at 2 in the morning. even isp's like plusnet are fine at 2am.

[Chrysalis]

ok I have just checked the post I made on VM's official forums.

they have confirmed a overload in my area, but apparently its in the final stages of been fixed already. I have asked them for a date on this and if its within a month I will be happy to wait otherwise I will ask for a discount and will be happy to pay full price after its fixed.

regarding my powerlevel the tech is due before midday so will report back if that improves things.

---------- Post added at 09:30 ---------- Previous post was at 08:09 ----------

ok tech been and gone

powerlevel fixed to 2.1db

peformance after he left is now in my sig for pingtest.

speedtest here

https://www.cableforum.co.uk/images/local/2010/04/84.png (http://www.speedtest.net)

He told me congestion is well known and its a 6-8 week fix. I am now in a situation whether to cancel or not as that is a very long time to fix. He also confirmed 50mbit has its own problems so moving to that wont bypass it.

[Ignitionnet]

Wasn't your area only just enabled as part of a new network build?

[Chrysalis]

it was yes. ignition is it still possible to hop between upstream channels like the old days?

I know if my downstream channel is maxed out I am screwed anyway. :( hoping its only an upstream problem.

[broadbandking]

There is only one upstream channel isn't there?

[Ignitionnet]

it was yes. ignition is it still possible to hop between upstream channels like the old days?

I know if my downstream channel is maxed out I am screwed anyway. :( hoping its only an upstream problem.

Yes it's possible to hop upstream channels still. Your issue isn't upstream congestion though I am quite sure of that I'm afraid.

---------- Post added at 15:46 ---------- Previous post was at 15:45 ----------

There is only one upstream channel isn't there?

On the overlay network there's one upstream per upstream service group yes, on the legacy uBRs still 3 - 4 upstreams per service group.

[Chrysalis]

yep confirmed to me by the ceo office today. I did hop anyway just to see if it would help, pings seem a bit more stable but speeds are actually a bit worse.

I will have more news next week, VM have acknowledged the problem which to be fair is good of them, I just feel if it takes months to install new equipment then they need to start upgrade work at a much earlier stage before things get this bad.

[Ignitionnet]

yep confirmed to me by the ceo office today. I did hop anyway just to see if it would help, pings seem a bit more stable but speeds are actually a bit worse.

I will have more news next week, VM have acknowledged the problem which to be fair is good of them, I just feel if it takes months to install new equipment then they need to start upgrade work at a much earlier stage before things get this bad.

Given it was new build there's no excuse for it happening in the first place to be honest.

Wonder why they can't just balance some 20M modems across the legacy and overlay networks given it's downstream congestion? The thought there is I guess that the overlay is a bit heavy with traffic as well.

Overall a total balls up though. Upgrade done on the cheap and nodes too big. Whichever bean counter forced the budget down so that the planner had to compromise should be beaten with a piece of coax.

[Chrysalis]

the tech told me the 50mbit is in a bad state, the official VM forums told me 50mbit is on the new network (overlay?)

so I am guessing that confirms your thoughts both are in a bad state.

I am wondering if this new build was done on the cheap and my area got put on to an existing area's ubr?

[Chrysalis]

guys an update.

at 3.30am a very quiet time of day.

https://www.cableforum.co.uk/images/local/2010/04/77.png (http://www.speedtest.net)

that was the best result out of 5, some were under 5mbit.

http://www.pingtest.net/result/14026342.png (http://www.pingtest.net)

again best result out of 5, it was much worse before I changed upstream channel.

If a upgrade on my UBR is splitting the users in half, I feel its so oversubscribed there would need to be 4 or 5 of these done at least.

actual download speeds are like this 2-5mbit

[Chrysalis]

I really dont know how VM can consider themselves able to get away with this, starting from 7am congestion was very noticeable and by 8am the connection was like dialup in terms of web browsing and unable to do anything heavy. Downloading of any kind of download site's had about 20kB/sec sustained throughput with numerous pauses.

Here is a link to my tbb graph which shows the extenct of loss and latency before I unplugged it.

http://www.thinkbroadband.com/ping/share-thumb/38b60e0c7f70398aff6894d709bb4709-05-04-2010.png (http://www.thinkbroadband.com/ping/share/38b60e0c7f70398aff6894d709bb4709-05-04-2010.html)

also some stuff from the modem log. I dont have timestamps for these.

SYNC Timing Synchronization failure - Failed to acquire FEC f...
ToD request sent- No Response received (multiple times)
Resetting the cable modem due to docsDevResetNow (multiple times)

[Chrysalis]

is there a way to check if its noise causing the problem? since these modems dont give any error stats, I always assumed cable was immune from noise but have recently read a few things that indicate some people have problems due to noise between them and their ubr.

[Sirius]

should be beaten with a piece of coax.

No thats not good enough, Lead pipe would be better.

[Chrysalis]

Well I last spoke to the ceo office on wednesday.

I think its likely I will quit VM even with the offer of free broadband.

I was told relief work was actually already done. :shocked:

There was an admittance clearly it wasnt adequate (ubr flatlined maxed out).

I then cant remember clearly what was said after that but I think it was something along the lines of a lot of heavy users in my area.

Was offered full credit for broadband but have to ring up every month to get it. So they seemed resigned that this problem isnt going away any time soon, probably cheaper to give me it free (and anyone else who complains) than to do major work.

When I pushed for a date, there is no date. Simply its going to be treated as urgent which 'may' mean only 2-3 weeks instead of months, also no idea if it would be more drip feeding of capacity or proper significant upgrades.

I asked how is 50mbit performance looking in my area, but couldnt get an answer immediatly and was supposed to phone me back the day after, its now saturday and no call so will try ringing them back monday.

[caph]

Chrysalis, I had the same problem a while ago and was told a relief date of November 2008. It finally got fixed in the middle of August 2009. You can take any "relief date" they tell you with a pinch of salt.

[Chrysalis]

caph sadly I cant disagree with you, I have used VM before (in its ntl days) and as a customer for 3 years congestion then was ongoing during the entire period as well except then it was less severe.

Maybe my area isnt fixable and just has too many heavy users so its abandoned so to speak in terms of upgrades and people who complain are compensated, who knows. But the level of congestion suggests a lot of UBR splits would be needed and not even one appears to be scheduled. Its all guess work at the end of the day but all I know is it isnt fit for purpose and I havent even bothered to use it for half a week now.

[Ignitionnet]

If you're in a student area or one with extremely heavy they just won't split it too much as the students, etc, will just nosh all the bandwidth. They'll just leave you all to get on with it.

It sucks, but there's no money in Virgin splitting the area down until it's a downstream per 60 customers which has been known to be required to mitigate congestion in some student areas. Not going to happen any more than any other operator would spend 10k on provisioning bandwidth for so few customers.

[Chrysalis]

well 60 users would only be about 15:1 contention, low but not that low. About 5 or so years ago their documented contention ratio I believe was something like 20:1 and my adsl service is 20:1 contention.

Thanks for the info and honesty, so realistically I cant expect this to be fixed. I dont know for sure if I am sharing with students but the nearest student area is not that far away, probably about 5-10 minutes walk.

I understand your point about the making money, but at the same time VM is arguably selling a service not fit for purpose, if they not willing to invest in capacity to mitigate heavy use then they should sell only lower speed services in areas such as mine, or capped usage services.

ISP's such as BE will also have their own fair share of students and they manage.

[Ignitionnet]

well 60 users would only be about 15:1 contention, low but not that low. About 5 or so years ago their documented contention ratio I believe was something like 20:1.

Thanks for the info and honesty, so realistically I cant expect this to be fixed. I dont know for sure if I am sharing with students but the nearest student area is not that far area, probably about 5-10 minutes walk.

I understand your point about the making money, but at the same time VM is arguably selling a service not fit for purpose, if they not willing to invest in capacity to mitigate heavy use then they should sell only lower speed services in areas such as mine, or capped usage services.

ISP's such as BE will also have their own fair share of students and they manage.

True, though they manage thanks to Telefonica being sugar daddy. If they had to pay their way they wouldn't come close and were having a fair few congestion issues before Telefonica gave them a few million quid to fix things.

The contention ratio is not going to make nice reading. I think I once calculated it in my area as being in excess of 100:1.

Think about this, 380 x 10Mbps modems on a 38Mbps downstream = 3800Mbps sold / 38Mbps provisioned = 100kbps provisioned per modem and contention ratio of 100:1.

[Chrysalis]

Ok so thanks to all who responded to my queries on this, ignition and seph especially.

I am curious what the 50mbit will be like so will try to see if I can use it but otherwise it probably means I will stick with adsl.

20kbit speeds at peak I think VM got a letter wrong in their ad ;)

[Chrysalis]

A new development, I turned on my TBB ping graph again but my ip changed so it turns out its pinging some other user on my UBR, their graph whilst still having jitter looks very significantly improved over how mine looked.

I have just now enabled a new graph on my current VM ip to see how they compare live.

The ip is my old one and is in same range so I am assuming the same UBR.

Would appreciate thoughts on what reasons would be another ip on the same UBR has significantly better latency and less variance from peak and off peak.

---------- Post added at 12:49 ---------- Previous post was at 12:26 ----------

confirmed they very different so me and another guy on the same UBR.

his min latency 20ms, average slightly above, max about 25-30ms.
mine min latency 20ms, average about 40ms (double) and max 100ms.

my line is idle.

---------- Post added at 12:58 ---------- Previous post was at 12:49 ----------

so here is 13 pings to my old ip now different user on same ubr and my current ip from one of my servers. The 3rd octet of ip also matches but censored it.

his

64 bytes from 86.14.old: icmp_seq=0 ttl=51 time=32.536 ms
64 bytes from 86.14.old: icmp_seq=1 ttl=51 time=31.159 ms
64 bytes from 86.14.old: icmp_seq=2 ttl=51 time=31.516 ms
64 bytes from 86.14.old: icmp_seq=3 ttl=51 time=30.125 ms
64 bytes from 86.14.old: icmp_seq=4 ttl=51 time=30.769 ms
64 bytes from 86.14.old: icmp_seq=5 ttl=51 time=34.786 ms
64 bytes from 86.14.old: icmp_seq=6 ttl=51 time=32.652 ms
64 bytes from 86.14.old: icmp_seq=7 ttl=51 time=31.190 ms
64 bytes from 86.14.old: icmp_seq=8 ttl=51 time=31.769 ms
64 bytes from 86.14.old: icmp_seq=9 ttl=51 time=32.426 ms
64 bytes from 86.14.old: icmp_seq=10 ttl=51 time=30.819 ms
64 bytes from 86.14.old: icmp_seq=11 ttl=51 time=31.365 ms
64 bytes from 86.14.old: icmp_seq=12 ttl=51 time=32.007 ms
64 bytes from 86.14.old: icmp_seq=13 ttl=51 time=30.376 ms

and mine (connection is idle)

64 bytes from 86.14.new: icmp_seq=0 ttl=51 time=61.776 ms
64 bytes from 86.14.new: icmp_seq=1 ttl=51 time=70.232 ms
64 bytes from 86.14.new: icmp_seq=2 ttl=51 time=79.434 ms
64 bytes from 86.14.new: icmp_seq=3 ttl=51 time=48.918 ms
64 bytes from 86.14.new: icmp_seq=4 ttl=51 time=39.584 ms
64 bytes from 86.14.new: icmp_seq=5 ttl=51 time=66.268 ms
64 bytes from 86.14.new: icmp_seq=6 ttl=51 time=33.564 ms
64 bytes from 86.14.new: icmp_seq=7 ttl=51 time=48.394 ms
64 bytes from 86.14.new: icmp_seq=8 ttl=51 time=36.631 ms
64 bytes from 86.14.new: icmp_seq=9 ttl=51 time=53.567 ms
64 bytes from 86.14.new: icmp_seq=10 ttl=51 time=71.819 ms
64 bytes from 86.14.new: icmp_seq=11 ttl=51 time=48.838 ms
64 bytes from 86.14.new: icmp_seq=12 ttl=51 time=60.342 ms
64 bytes from 86.14.new: icmp_seq=13 ttl=51 time=44.689 ms

---------- Post added at 13:45 ---------- Previous post was at 12:58 ----------

ok some insight in this would be good, it seems I cannot get off upstream channel 2, 3 is hard but possible but the other 2 upstream channels appear to be locked off.

I pinged about 50 other ip's on the same UBR and about half werent pingable (firewall, offline etc.) about 20 were good pings and about 5 have my problem.

My guess is right now is upstream channels 2 and 3 are bad, the other 2 or one of them is good but my modem wont connect to them, is possible I am onto something here? or is it possible the good ip's are 50meg users who same the same UBR?

[Chrysalis]

Thanks to the lady who has helped me in the CEO office, I am positively surprised she has managed to get things much better for me.

Last night I started using my VM connection again and it appears I got moved to a different UBR port. (overlay network)

When tested last night at 5pm sunday and still the same at 6pm I was able to get 19-20meg throughput on both speedtest.net and a single stream ftp test.

I have extra base latency now which I guess means the port I am connected to is some distance away, but is a minor issue I will eventually get used to. Jitter is still apparent but is nowhere near as severe now.

9ms jitter on pingtest.net just now which doesnt seem great for 5am but I also got 10ms yesterday at 5pm which for peak time is ok.

So I will praise VM when due, not mr negativity like some guys have said :)

Hopefully now this will stay better and I wont have reason to moan again.

Ultimately my old UBR port I think should be fixed, and I hope it does get fixed but at least they found a solution to my connection before that gets done.

[Chrysalis]

ok an update, in under 2 months after been moved to the overlay channel on my port there is defenite significant degrade.

However the degrade is not in speeds, I still get speeds around the 20mbit mark for the majority of the time. The problem is latency stability is now getting bad again and latency sensitive stuff such as ssh/telnet is hard to use.

Is there any scope of complaint here? as I understand it my problem is similiar to gamers who also require good jitter and that if speeds are good VM consider there to be no problem?

Obviously I am also concerned that with incoming upload speeds on the service even with docsis2 upgrades latency is likely to get worse. Ignition hinting that they will add extra channels if congestion increases but if thats a policy why arent they doing that now?

[Ignitionnet]

Obviously I am also concerned that with incoming upload speeds on the service even with docsis2 upgrades latency is likely to get worse. Ignition hinting that they will add extra channels if congestion increases but if thats a policy why arent they doing that now?

Per el blog - congestion is resolved by increasing the bandwidth available per customer. This is done by:

A) Increasing available bandwidth to a group of customers and/or
B) Reducing the customers sharing each unit of bandwidth.

A) Is done by upgrading upstreams to DOCSIS 2.0 or 3 standard, B is done by

1) Segmenting return path to reduce number of customers per upstream service group and/or
2) Adding another upstream to the existing upstream service group.

VM's approach is usually B/1 with some A and the very occasional bit of B/2.