50M Why is latency & jitter so bad on DOCSIS3?

ianch99 · 17-09-2012, 20:07

I am using the 50M product connected to the sotn8 CMTS. For the last few months the latency and jitter has been variable: really bad during the evening, less so during the day. There is a utilisation fault ticket logged (F002097435) that was supposed to be fixed a month or two back but then was bounced to 12th Sep which I had assumed was the cause of the poor TBB graphs I see.

However ... my neighbour is on the 10M product: DOCSIS2 w/a Media Hub and his TBB graph is perfect. My graph and another neighbour's (who is also on the 10M product *but* using a SuperHub in modem mode) are terrible ..

.. and here is the conundrum .. all three of us are connected to the same CMTS: sotn8

Why would a DOCSIS2 connection using the same physical routing to the same CMTS have such a different latency and jitter profile?

Here's a snapshot of today's TBB:

graphs for the 3 connections:

kwikbreaks · 17-09-2012, 20:14

What channels do the various connections use? I'm guessing that the good one is different from the bad ones and it's that rather than docsis2/3 that counts.

craigj2k12 · 17-09-2012, 20:49

the 10mb on the docsis1.1 network will just have lower upstream utilisation

the 50mb on docsis3 is on a heavily congested upstream channel

ianch99 · 17-09-2012, 21:11

Two good answers .. thank you. I would need to check the frequencies but I suspect you are right.

But why would the CMTS device have such a segregated internal architecture such that different frequencies and or different cable modem protocols result in such in imbalance of network capacity allocation?

Surely as time goes by, more of the VM customer base will be switching to DOCIS3 devices (mainly Super Hubs) leaving fewer and fewer "legacy" (DOCSIS 1/2) devices on the network.

How can it be that within the same CMTS, you have a minority of legacy customers enjoying good latency and jitter (and so benefiting from good Game play and VOIP service) and the remaining majority of DOCSIS 3 customers suffering from high latency & jitter with resulting bad online Gaming and VOIP quality?

What we see here is: "pay more, get less" for DOCSIS 3 customers and "pay less, get more" for the legacy DOCSIS 1/2 customers ..

It is ironic that the friend who was very happy on the 10M legacy product was "upgraded" to the Super Hub when her modem broke. Since then, although she has the same speed as before, her online experience is much the worse esp. for her Skype calls

MaverickJesus · 17-09-2012, 21:51

In this case its that way round - in my case (stav) is the opposite. I've seen the utilisation graphs and the DOCSIS1/2 segment is so bad its obscene, whereas my 60mb is fine.

Chrysalis · 17-09-2012, 23:36

I dont think docsis3 by nature has worse latency than docsis1 just that those graphs you posted the 2 docsis3 ones are on oversubbed channels.

When I was on docsis1 it was extremely bad, graphs filled up with yellow almost 24/7.

also its possible the good neighbour is on a different segment, my neighbour is on a different segment to me.

---------- Post added at 23:36 ---------- Previous post was at 23:33 ----------

Quote:

Originally Posted by ianch99

Two good answers .. thank you. I would need to check the frequencies but I suspect you are right.

But why would the CMTS device have such a segregated internal architecture such that different frequencies and or different cable modem protocols result in such in imbalance of network capacity allocation?

Many of us have been waiting for the answer to this question for years, only VM know why they balance different segments so imbalanced.

qasdfdsaq · 18-09-2012, 17:18

They're not all that unbalanced, the charts I've seen show them to be fairly close in terms of subscriber numbers, it's the unpredictable usage patterns that are the problem - it only takes one 100/10 user uploading at full whack to create noticeable jitter for 150+ other customers... God forbid two of them hop on a busy torrent at the same time and you've got complete network saturation.

Chrysalis · 18-09-2012, 23:28

It can be made more predictable by limiting the usage :p

But yes you have pointed out why what they doing is so irresponsible.

If you sell cheap fast unlimited broadband at low cost I wouldnt say its too hard to predict some people will cane the connection. Especially when the top package also has no STM. Capacity planning should assume at least 5% of the customers in a segment will use it 24/7.

Eeeps · 19-09-2012, 14:54

Quote:

Originally Posted by ianch99

Surely as time goes by, more of the VM customer base will be switching to DOCIS3 devices (mainly Super Hubs) leaving fewer and fewer "legacy" (DOCSIS 1/2) devices on the network.

I think you have answered your own question there. As customers are migrated to DOCSIS3 from 1/2, the contention on the channel(s) used for DOCSIS 1/2 is reduced.

I guess that eventually there are so few DOCSIS 1/2 customers on a segment that VM disconnect the channel. The customers ring in to say their BB is not working and VM say that their modem is broken an they will need a replacement (SH). The channel can then be re-allocated for DOCSIS 3.

Sephiroth · 19-09-2012, 20:43

[QUOTE=Chrysalis;35475314]I dont think docsis3 by nature has worse latency than docsis1 just that those graphs you posted the 2 docsis3 ones are on oversubbed channels.

When I was on docsis1 it was extremely bad, graphs filled up with yellow almost 24/7.

also its possible the good neighbour is on a different segment, my neighbour is on a different segment to me.

---------- Post added at 23:36 ---------- Previous post was at 23:33 ----------

Many of us have been waiting for the answer to this question for years, only VM know why they balance different segments so imbalanced.[/QUOTE]

Who says it's a different segment? The neighbour would be on the same segment. It's a matter of shared frequencies and the DOCSIS 1 user is on an undersubscribed frequency.

ianch99 · 19-09-2012, 22:25

Ok dumb question: why does a cable modem's RF frequency used to communicate with the CMTS have any bearing on the (IP) network bandwidth available on the network "beyond" the CMTS? .. or is this too simplistic?

Sephiroth · 19-09-2012, 22:52

Quote:

Originally Posted by ianch99

Ok dumb question: why does a cable modem's RF frequency used to communicate with the CMTS have any bearing on the (IP) network bandwidth available on the network "beyond" the CMTS? .. or is this too simplistic?

No bearing at all. The network beyond the CMTS can get crowded of course - at many points including the internet. But I don't understand your point.

I answered the question posed in your first post.