The FairShare system, which works with Sandvine’s network switches, allows service providers to curtail bandwidth based on subscriber-usage metrics from various sources to balance available bandwidth and resources among all subscribers.

Donnelly said FairShare has already been deployed by a North American service provider, but he declined to identify the customer.

FairShare uses technology from two companies Sandvine acquired last year, CableMatrix Technologies and Simplicita Software.

said by jc100 :

First, this software I am sure is deep packet inspection. IT HAS TO LOOK AT what you download or for certain protocols to determine it's p2p. (Invasion of privacy).

said by en102 :

Yup - this will be interesting

Throttle by user bandwidth for long periods of time may not be a good idea either.
Eg. I'm on a 3Mbps connection and I want to d/l Fedora Core DVD ISO as well as some other ISO's. A reset would just tick me off.

Why not look at the overall traffic crossing the node, and when its nearing capacity, throttle from the top usage down until its at a supportable level.

said by TK Junk Mail :

The easiest way would be to throttle down the transfer speed at the CMTS for those users whose traffic is staying at the highest rates for lengthy periods of time.

said by halfband :

Isn't that similar to what the satellite guys do now, throttle a high volume user? They refer to it as FAP right?

said by TK Junk Mail :

The real question is how do they then throttle the traffic after identifying whose packets need throttling. The easiest way would be to throttle down the transfer speed at the CMTS for those users whose traffic is staying at the highest rates for lengthy periods of time.

Throwing resets into the stream would be a way that will still get people upset about forging packets.

said by TK Junk Mail :

Yes, I believe it is the same idea. Though the way I read Fairshare is that it is more dynamic than a satellite FAP system in that it will be applied and then removed real time and much more quickly and for a very limited period of time - maybe for just the length of a download or upload.

Capital expenditures for our businesses were $781.3 million, $885.8 million and $768.7 million, in 2007, 2006 and 2005, respectively, and primarily include payments for consumer premises equipment, such as new digital video cable boxes and modems, as well as infrastructure and capital expenditures related to our cable and Lightpath telecommunications networks

an increase in Cable segment capital expenditures of 41.2% to approximately $6.0 billion, primarily as a result of (i) the installation of advanced set-top boxes, modems and other equipment associated with the increase in subscribers to our digital video, high-speed Internet and digital phone services; (ii) network improvements to handle the growth in subscribers and to provide service improvements and enhancements

Total capital expenditures for the years ended December 31, 2007, 2006, and 2005 were approximately $1.2 billion, $1.1 billion, and $1.1 billion, respectively.

$17,717[M] in construction and capital expenditures

We continue to make significant capital expenditures to meet the demand for communications services and to further improve such services. Capital spending for Wireline was $10,956 million in 2007, $10,259 million in 2006 and $8,267 million in 2005. Capital spending for Domestic Wireless was $6,503 million in 2007, $6,618 million in 2006 and $6,484 million in 2005.

said by TK Junk Mail :

Throwing resets into the stream would be a way that will still get people upset about forging packets.

said by espaeth :

I've skimmed through the Cisco CMTS config guides and I don't see any obvious way to implement equal division traffic shaping without resorting to a 3rd party box.

said by espaeth :

The TCP reset trick is limited to a small group of applications that can gracefully deal with a lost connection and can still assemble working data at the end of the transfer. P2P apps do this very well by not strongly relying on any one particular connection and CRC checking the received file to ensure a valid copy was reached.

said by funchords :

Why would you want to? Every modem is already configured to limit uploads. The customers' traffic is already shaped.

You only need to shape it again if you've oversold the bandwidth beyond all reasonable ability to deliver it.

said by espaeth :

That's a very narrow view of the situation; you're assuming continuous data transfer which is atypical in a world of burst traffic.

said by funchords :

Yeah, which is why I couldn't upload any of my original content for three straight months, 24 hours a day, 7 days a week.

said by funchords :

Repeating the same mistake twice is beyond dumb.