50M Think Broadband Ping Monitor Results (POST YOURS) - Page 54

kwikbreaks · 22-12-2011, 10:41

Of course now I'm only 10/1Mbps it has to get pretty bad before I see speed reductions if things work how I think they do. I'm just going to sit things out on this package now pretty much for sure as Infinity has been brought forward by 6 months now.

Who knows 7 months from now I might be berating BT and hankering on returning to VM.

Chrysalis · 22-12-2011, 17:26

yep as far as I know there is no priority given to higher tier customers under congestion so a 100mbit customer could be seeing 11mbit speeds a near 90% drop whilst a 10mbit customer could still see full speed.

kwikbreaks · 22-12-2011, 17:49

What a terrible shame

I spotted I couldn't logmein to home today and was rather hoping that this outage like the last signalled a node split. No such luck it just signalled an outage - at least it wasn't a dipstick engineer swapping the tap to some other random one this time though which is a minor consolation.

Chrysalis · 23-12-2011, 10:58

that outage may still have been reseg work, but you still on the original segment.

Chrysalis · 23-12-2011, 16:20

connection going to hell again here.

djfunkdup · 23-12-2011, 16:26

Quote:

Originally Posted by Chrysalis

yep as far as I know there is no priority given to higher tier customers under congestion so a 100mbit customer could be seeing 11mbit speeds a near 90% drop whilst a 10mbit customer could still see full speed.

i don't think that is correct....i get 90-100 Mbps 247

even when the local UBR is loaded @ 42% i can still hit these speeds

Chrysalis · 23-12-2011, 17:18

Quote:

Originally Posted by djfunkdup

i don't think that is correct....i get 90-100 Mbps 247

even when the local UBR is loaded @ 42% i can still hit these speeds

42% is low for VM.

So the fact your UBR is relatively uncongested means what to my statement?

craigj2k12 · 23-12-2011, 20:09

with 200mbit of available bandwidth, the uBR would need to be over 50% load before you would see any speed reduction, so at 42% your going to be hitting full speed 24/7

kwikbreaks · 24-12-2011, 10:02

I guess it all depends at what level you are measuring. The 200Mbps is at the node level (or even sub node? - depends on definitions) and the modems served will be in the low (hopefully very low) hundreds. You could have a completely borked node yet the overall CMTS (which I think serve several thousand modems) could be at a tiny overall utilisation level.

That is a couple of torrent freaks nearby may wipe out your street but they will have no impact at all on the overall district. This is why cable is more susceptible to heavy usage than xDSL where the contention point is thousands of users on far fatter pipes.

Unless of course (as is often the case) I've got it all wrong.

Chrysalis · 24-12-2011, 14:09

Quote:

Originally Posted by kwikbreaks

I guess it all depends at what level you are measuring. The 200Mbps is at the node level (or even sub node? - depends on definitions) and the modems served will be in the low (hopefully very low) hundreds. You could have a completely borked node yet the overall CMTS (which I think serve several thousand modems) could be at a tiny overall utilisation level.

That is a couple of torrent freaks nearby may wipe out your street but they will have no impact at all on the overall district. This is why cable is more susceptible to heavy usage than xDSL where the contention point is thousands of users on far fatter pipes.

Unless of course (as is often the case) I've got it all wrong.

Thats possible as I understand it.

eg. leics only had 1 ubr covering the entire city so that ubr would be serving several nodes at the very least. The situation described probably does exist in some areas and makes a mockery of what VM are doing since clearly things can be balanced better but are not.

Ignitionnet · 24-12-2011, 15:39

Quote:

Originally Posted by kwikbreaks

I guess it all depends at what level you are measuring. The 200Mbps is at the node level (or even sub node? - depends on definitions) and the modems served will be in the low (hopefully very low) hundreds. You could have a completely borked node yet the overall CMTS (which I think serve several thousand modems) could be at a tiny overall utilisation level.

That is a couple of torrent freaks nearby may wipe out your street but they will have no impact at all on the overall district. This is why cable is more susceptible to heavy usage than xDSL where the contention point is thousands of users on far fatter pipes.

Unless of course (as is often the case) I've got it all wrong.

That's fine, though there is no 'sub node'. The node is as far as it goes and there's no way to break modems down into smaller groups apart from adding more channels and spreading devices across them.

Each 10k can hold 8 cable line cards, one line card that comes to mind can have 72 downstreams and 60 upstreams, split that 12 ways you've 12 service groups of 6 downstreams and 5 upstreams.

Basically fully load a Cisco 10k with that you've close to 24Gbps of downstream and, using the current schemes VM are, 8.7Gbps of upstream capacity.

craigj2k12 · 24-12-2011, 15:46

Quote:

Originally Posted by Ignitionnet

Each 10k can hold 8 cable line cards, one line card that comes to mind can have 72 downstreams and 60 upstreams, split that 12 ways you've 12 service groups of 6 downstreams and 5 upstreams.

Basically fully load a Cisco 10k with that you've close to 24Gbps of downstream and, using the current schemes VM are, 8.7Gbps of upstream capacity.

So what are VM using, if they have this many upstream channels available why do they all seem so congested? Most areas are using 4 downstreams and 1 upstream, and if theres 72 down and 60 up, why cant it be 4 down and 3 up for each modem, that would certainly lower ping and jitter and stop TBB ping graphs looking like a house fire.

Im sure theres a reason why not, but I cant work it out myself

Ignitionnet · 24-12-2011, 15:59

Quote:

Originally Posted by Chrysalis

Thats possible as I understand it.

eg. leics only had 1 ubr covering the entire city so that ubr would be serving several nodes at the very least. The situation described probably does exist in some areas and makes a mockery of what VM are doing since clearly things can be balanced better but are not.

Given there are 17 of them there in Leicester's own hub site now, 4 of which are Motorola BSRs VM must have been pretty busy getting the other 16 onto the network in such a short time, and of course there is also the minor issue that Leicester is served by more than one hub site, so VM must've done something pretty special and made Leicester pretty unique for a city served by Leicester and Northfields (which has 12 VXRs and 2x10k) to all be hooked to a single uBR.

I must admit I struggle to see exactly why VM would put 2 Motorola BSRs into my hub with its 4 VXRs, one of which was only half full, and only use 1 for a site with 13 VXRs.

If space were a problem they'd either wait until more space was available or do a displacement build, swapping VXRs for BSRs directly.

That was well worth clicking the 'View Post' button for.

---------- Post added at 15:59 ---------- Previous post was at 15:57 ----------

Quote:

Originally Posted by craigj2k12

So what are VM using, if they have this many upstream channels available why do they all seem so congested? Most areas are using 4 downstreams and 1 upstream, and if theres 72 down and 60 up, why cant it be 4 down and 3 up for each modem, that would certainly lower ping and jitter and stop TBB ping graphs looking like a house fire.

Im sure theres a reason why not, but I cant work it out myself

Most areas with any kind of load are using multiple upstreams, in addition most areas have more than one node sharing the downstream, say the 4 downstreams split across 2 nodes, with each node having 2 independent upstreams.

Very few areas running on a single DOCSIS 2 upstream, majority of issues are down to 100Mb users going nuts and bad congestion handling by the Cisco 10k. You can tell a congested Motorola from a congested 10k pretty easily.

Once line card swaps are finished VM can do upstream bonding.

kwikbreaks · 24-12-2011, 16:18

Quote:

Originally Posted by Ignitionnet

The node is as far as it goes and there's no way to break modems down into smaller groups apart from adding more channels and spreading devices across them.

That's what I was wondering about when I invented the term sub-node - I wondered if a "node" whatever it may be could have multiple channel groupings giving it more than 200/18.

Ignitionnet · 24-12-2011, 18:15

Quote:

Originally Posted by kwikbreaks

That's what I was wondering about when I invented the term sub-node - I wondered if a "node" whatever it may be could have multiple channel groupings giving it more than 200/18.

Yes they could but there's no point, far better to bond all the channels together.

A node however is a physical construct, what you're talking about would be called a service group.