how long till 1gbit?

[Andrewcrawford23]

Originally Posted by qasdfdsaq

I thought they were already aiming at a 250-homes-per-node target

im not sure i know it somewher ebetwen 200-500 i think it depend son areas hence some are worse than others

[craigj2k12]

yeah some areas are worse off in that respect but qas is talking about the target number per node

[Sephiroth]

Originally Posted by General Maximus

Just had a quick scan of the news and saw this:

http://www.foxnews.com/scitech/2012/...ntcmp=features

I would love to have 1gbit for the sake of it but despite how much I download (about 400gb/month) I can honestly say I just wouldn't use it to it's full potential and it would sit there doing nothing 99% of the time.

We can all dream. Hopefully we'll be getting close to it within the next 10 years. It is going to be something when we can sit here and remember what it was like back on 33k dialup and then say we have now got gigabit. It will be like snail speed compared to a Concorde.

Coming back down to earth from the general discussion so far in this thread, the FOX article is very incisive. It points out that 1Gb/s downloads are just not to be expected. My remarks below are confined to the VM network.

The reason for this, apart from the rate limiting or whatever is used by a distributor to control demand, pretty well guarantees that there's little premium value for downloads in a 1 Gb/s service.

The second point that could well perplex the likes of VM would be peering arrangements and the costs of raising the bandwidth to keep customers happy. We all know that VM keep these costs (and hence bandwidth) to the minimum they can get away with.

Third, there are potential consequences for VM's core since each connection would be able to punt in at 10 x what they can do now (subject to congestion at the oprical node LOL).

Finally on VM's 1.5 Gb/s trial - I haven't been able to glean what they've actually done. They say it's on a standard infrastructure; but they haven't disclosed the QAM number nor any amount of detail that would make it worth knowing that they've done this.

[qasdfdsaq]

Originally Posted by Sephiroth

The reason for this, apart from the rate limiting or whatever is used by a distributor to control demand, pretty well guarantees that there's little premium value for downloads in a 1 Gb/s service.

This reminds me of something I'd thought of as well - the only likely way for any one user to take up the full 1Gbps capacity would be the likes of BitTorrent - which downloads simultaneously from many different hosts in many different locations in the way HTTP downloads, by default, do not. I do wonder then, if availability of such connections would likely cause a resurgence in the level of P2P use.

[Andrewcrawford23]

Originally Posted by qasdfdsaq

This reminds me of something I'd thought of as well - the only likely way for any one user to take up the full 1Gbps capacity would be the likes of BitTorrent - which downloads simultaneously from many different hosts in many different locations in the way HTTP downloads, by default, do not. I do wonder then, if availability of such connections would likely cause a resurgence in the level of P2P use.

acutally a lot of http downloads now a days use a p2p model, not that they use the user downloading to upload in more that says for a linux ditro ther eis multiply host for the file they have some sort of program or code set up that when you start downloading it it then connects to peers

dnt ask me how they do it as a http download should only ever use one conenction adn oyu need a download manager to use more than one host but they seem to have developed something server side that does the same

so if that sort of things is rolled out with majority of websites then 1gbit could be archive easily

i cant remebre the site that i found that was doign this multi conneciton http download but it wsnt a linux distro it was jsuta example

how did i find out it was doing it i was packet capturing data at th etime looking for some information sinc ei suspect i might have a trojen on my system at the time

[qasdfdsaq]

Originally Posted by Andrewcrawford23

acutally a lot of http downloads now a days use a p2p model, not that they use the user downloading to upload in more that says for a linux ditro ther eis multiply host for the file they have some sort of program or code set up that when you start downloading it it then connects to peers

dnt ask me how they do it as a http download should only ever use one conenction adn oyu need a download manager to use more than one host but they seem to have developed something server side that does the same

so if that sort of things is rolled out with majority of websites then 1gbit could be archive easily

You might be thinking of load-balancers. They're nothing like P2P downloads and have no effect on the actual connections between the server and the end user, only what invisibly goes on behind the server.

Short of a download manager, what you're saying isn't possible

[Andrewcrawford23]

Originally Posted by qasdfdsaq

You might be thinking of load-balancers. They're nothing like P2P downloads and have no effect on the actual connections between the server and the end user, only what invisibly goes on behind the server.

Short of a download manager, what you're saying isn't possible

nothing is impossible, but i do take your point it most like aloa blanacer, , acutally in thoery i could possible scritp code that do excately wha ti said above but it then make the server a client so it would be downloading from multiply sources and passing that speed to you as i say in thoery how i script that in pratice i aint sure not really lookin to do it not unless i am paided

[General Maximus]

Originally Posted by Sephiroth

They say it's on a standard infrastructure; but they haven't disclosed the QAM number nor any amount of detail that would make it worth knowing that they've done this.

I have been reading more about QAM and trying to understand what it is and how it works. I found a paper last night (and I cant find the link for it now) which was discussing alsorts of stuff to do with the problems with noise, snr and how you would calculate BER (can't remember what that is either).

The thing which caught my eye and why i was looking to start off with is how you calculate the throughput on a line for each QAM number. They gave 2 examples:

QAM64 = 2^6 = 64bits. 6 (the power) * 5 = 30mbits per channel
QAM256 = 2^8 = 256bits. 8 (the power) * 5 = 40mbits per channel

I am a bit confused because I don't understand why we are multiplying the power by "5" and for whatever reason I would have thought that it should be more around 7 because I thought QAM256 was 55mbits per channel and not 40.

I await your words of wisdom Seph

[rickymallory]

^^ most likely those figures are for the DOCSIS they use in the US and not EU DOCSIS which we use here. DOCSIS has less bandwidth than EU DOCSIS.

DOCSIS = 42.88Mb/s per channel
EU DOCSIS = 55.62Mb/s per channel

[General Maximus]

cool, so I was right with the x*7 then.

That means for 1gbit we would be looking at QAM2048 for 1gbit and QAM4096 for 1.5gbit. Now I just need to understand how SNR comes into it

[Andrewcrawford23]

Originally Posted by General Maximus

cool, so I was right with the x*7 then.

That means for 1gbit we would be looking at QAM2048 for 1gbit and QAM4096 for 1.5gbit. Now I just need to understand how SNR comes into it

if you are using shub as a exmplae then yes, but to go to such qam would reuqire a much strict torlance in upstream and downstream power levels i cant remember what extcately in fac tno sute that qam4096 exist

to achive 1.5gbit virign used a modem capable of bonding up to 32 channels so 32*55mbit but comcast i think have tested a 64 bonding one and i think tested at qam1024i think they achive about 4gbit

[General Maximus]

I didn't think about that. It is probably easier for them just to stay on QAM256 now that they have got the network sorted for capacity and power levels etc and just bond more and more channels. They could offer up to 800mbit with a 16 channel modem.

[Andrewcrawford23]

not saying it impossible they wont go beyond qam2546 as i have noticed a lot fo work that has improve power levle sin my area drastically my area had a issue with upstream stuck beyond 58-61 they done work at the headend and brought it down to about 53 now they have done mroe owkr and brought it down to 40, the downstream has been changinga lot recently to from 6 to -8 now about -2

[qasdfdsaq]

Going beyond QAM256 is indeed, a lot of work for very little gain, and well into the realm of diminishing returns. Very few public/consumer networks use anything above QAM256.

Going from QAM256 to 1024 for example requires your equipment to be four times more precise while only giving you 25% more throughput. QAM4096 requires 400% better precision again, with only 20% gain.

[General Maximus]

maybe they could do a bit of both and go to something like QAM512 and 16 channels