Comcast's Approach To Congestion
 

Little something form the largest cableco in the world:

http://downloads.comcast.net/docs/Co...i-20080528.pdf

http://www.comcast.net/terms/network/

I'll summarise differences between this and STM

There is no 'throttling' just priority, customers who are using heavily will have a lower priority than less heavy users.

Unless the network ports are nearing congestion there is no prioritisation at all. Unless the network actually needs the bandwidth control to avoid bandwidth issues it doesn't happen.

In addition the following being done to assist with P2P usage:

– Tracker optimizations (optimizing & localizing P2P flows)
– Caching
– P2P client optimizations

This strikes me as a better approach. The only time that any prioritisation happens is when the network is nearing congestion. This seems to me to be far closer to 'preserving the customer experience' than the VM approach of throttling you down whether your area is congested or not and whether you are affecting other customers or not.

Also I like the idea of just prioritising traffic, so that customers who are heavily using data can continue to use whatever bandwidth is left over from customers using less. To me anyway that is preserving the experience of lighter users while giving heavier ones whatever resources are left. If the bandwidth is there why not use it?

Just for the sake of inflicting pain on us subs / soon to be subs

A sub with 384Kbps upstream will go to 1Mbps
A sub with 768Kbps upstream will go to 2Mbps

Oh for 2Mbps on cable.

Downside being it takes a bit more than a couple of lines of config on a uBR, but sure if Virgin are more interested in the customer experience than saving money you'd think they'd have no problems implementing this somewhat more granular approach :)

Any thoughts?

Re: Comcast's Approach To Congestion
 

Simple reason virign then need ot pay for the bandwidht outside there network and at the end of the day that why they stm so they reduce costs

Re: Comcast's Approach To Congestion
 

Also remember that this new throttling is a result of an FCC ruling against Comcast and a large fine imposed for its previous secret throttling methods.

Re: Comcast's Approach To Congestion
 

Yep, which is nice.

Talking of which Frank I note you left Rogers, now there was a hardcore company in the field of shafting their customers. Hard caps, overage charges and P2P throttling. :)

Re: Comcast's Approach To Congestion
 

You might like to read TorrentFreak’s Uncovering The Dark Side of P4P. While TorrentFreak has a certain bias, the concerns raised about monitoring and control are valid.

Comcast will not be fined.

Re: Comcast's Approach To Congestion
 

Please could this be merged with http://www.cableforum.co.uk/board/63...fairer-stm.htm

Caching and 'tracker optimisation' has been happening for a while, P2P flows being redirected to on-net peers to save transit bandwidth. Potentially while P4P makes some intrusion easier it's fairly trivial still for ISPs to poke into P2P flows, especially with the sub-standard encryption that's employed in such clients. Worst case ISPs could even proxy the encryption transactions if they had to.

This is quite an interesting technology actually, I might have to do some reading into it and educate myself a bit :)

Re: Comcast's Approach To Congestion
 

But STM doesn't stop me downloading, I still use (legal) torrents, I still share and watch video on YouTube, I still download HD movie trailers etc.

They're still having to pay for the bandwidth as STM doesn't make me think about the way I use bandwidth, it just gets to me slower.

The way we use the internet has changed from what we did 10 years ago, media rich online learning, high bandwidth AV content, purchasing music and films online, software distribution and all the others things we now do.

VM have to evolve with this and increase the bandwidth available at their peering points not STM their customers. How they afford this is up to them, if they can't afford to do it because of the ridiculously low broadband prices they offer people they shouldn't be doing that.

Wholesale IP transit charges have come down in the last 10 years, not gone up, yes we're using more of it but I was still paying £30 a month 10 years ago for a 33kbit/s connection and using less bandwidth. Where has the investment been?

Re: Comcast's Approach To Congestion
 

It saves bandwidth as ISPs pay for 95th percentile usage.

Reduce the peak traffic load through STM and you reduce the 95th percentile and therefore the bandwidth bill.

EDIT: On VM's network however and most other ISPs the cost of the transit and peering will not be anywhere near as high as the cost of the access network bandwidth. A few quid a Mbit for transit versus the cost of optics, a CMTS card, fibre and lasers for 38Mbit/s makes the transit not that much of an issue.

Re: Comcast's Approach To Congestion
 

From the Comcast document.

And VM are increasing available bandwidth by introducing 50meg connections on DOCSIS 3

Some form of traffic management is required, exactly what form and at what expense is the question. VM's current model sucks and I think Comcast's is certainly an improvement, though the best solution? I don't know. If you make any attempt to prioritse real time traffic, whats to stop someone writing P2P software that uses VOIP protocols (for instance) and pretends to be something it isn't?

Re: Comcast's Approach To Congestion
 

Bandwidth (at least how VM measure it) is actually the amount of data being carried at ony one time. It is not the total amount of data carried. If you are using (say) 20 Meg, you are using that amount of bandwidth whether you download a 1 meg file or a 1,000 meg file. You are just using it for longer if you are download a larger file.

They aren't trying to stop people downloading (if they were, they'd have hard caps, so your download stops at a certain point, or they'd charge for downloads above this point). What they are trying to do is reduce the strain on the network of people downloading.

I am not saying whether that way is right or wrong though.
They are suffering (now) from a lack of foresight years ago. When I first started using the net in 1994, there were small signs that this sort of stuff (streaming TV, music downloads). Admittedly, nothing specific, but there were signs that broadcasting was coming.

There *is* a technology available that if enabled, would reduce a lot of bandwidth requirements (at least for live streams). Multicasting. The standard way of transmission on the Internet requires a connection from the server to the user, so the server may end up dealing with 10,000 users (and use a lot of bandwidth). This is called Unicasting. Multicasting uses routers along the way to cache and distribute the data. So, if an ISP has (say) 10,000 people watching a live stream, with Unicasting, they'll have 10,000 copies of the same data travelling over their links to the Internet. If they use Multicasting, their own routers will handle distribution of the data, and they'll only have a few copies of the data travelling over their link.

The trouble is, Multicasting only works for live streams, and only works if every network the data travels through is multicast enabled. As such, it is unlikely to happen.

What do you base the assertion that wholesale transit charges have gone down on? The broadband ISP market (in this country anyway) has traditionally run on loans, venture capital, cross subsidy from other products and loss leading products, so the prices we are charged probably bear little resemblance to what it actually costs for any ISP to serve us.

Re: Comcast's Approach To Congestion
 

Multicast wouldn't help with the major congestion on VM's network, the connection between the uBR and the customers though.

Wholesale transit costs have gone down greatly. You can pay less than £10/Mbps/month now without buying in the sort of quantities a multi-million customer ISP would, it's not that many years ago that three times that was an excellent price. Internet bandwidth has never been cheaper.

http://www.dslreports.com/shownews/95203

£2/month per Mbps on a 10GbE port with a 3 year contract in the states. Just a Google search for transit shows under £10/Mbps/month in the UK from resellers of transit.

Re: Comcast's Approach To Congestion
 

it's a shame that £2/month requires consuming a whole 10GigE port which i dont many users would do :p: and the fact it's cogent would put me off massively, there is a reason they're that cheap.

Re: Comcast's Approach To Congestion
 

Good reason why the news story called it McBandwidth :)

Re: Comcast's Approach To Congestion
 

Now multiply that £10mps by 20 Meg. Assuming VM get some sort of bulk discount, we can be generous and assume it costs them £100pcm to give you your 20 meg. That's a 1:1 connection (which admittedly, VM don't offer). This is why they use many to 1 contention (they state the average is 20:1), so that cost can be divided up amongst many customers.

Their business model (along with that of most ISPs) is based on the assumption that most of the users will use hardly any of their available bandwidth. From what I have been told, VM take quite heavy losses on their top-tier product, and make up these losses using the lower tiers.

Add in things like iPlayer (or any web based VOD service, even Youtube), and all of a sudden, the people who rarely bothered using their connection to it's full potential start to do so more often. Too many of these people do this, and the business model falls apart.

Unfortunately, it is not really practical for ISPs to offer broadband for the low prices it is offered at now, unless they can subsidise the cost with another product, use STM or Cap the service.

Re: Comcast's Approach To Congestion
 

I, personally, don't have a problem with STM as such. Yes, I think the limits could be slightly more generous, but I do think ANY ISPs network needs some sort of traffic shaping on it. ISPs, despite what many users may think, don't have an infinite amount of cash to splash on peering/bandwidth arrangements.

I think though, that VMs pricing structure is slightly skewed... In the way that people on lower tiers are subsidising the higher tiers. My question is WHY? People, if they want faster BB should PAY!

I'm only on 2mb and I pay 18/month right now. So much in my opinion (as opposed to some providers) that I'm considering changing phone/bb to accomodate this. Afterall, I could get 16mb for £10/month (on O2 LLU). I'm not a heavy downloader so I don't worry about STM/Traffic Shaping. Theres no need for the heavy users to moan because they should be paying much more for the service they currently get.