dslreports logo
 
    All Forums Hot Topics Gallery
spc
uniqs
34

TZi
join:2001-07-05
Miami Beach, FL

TZi to Corydon

Member

to Corydon

Re: Don't assume anything

said by Corydon:

The main problem I have with Comcast is how opaque their practices are. They really need to be a lot more open about what they're doing.
I too believe that Comcast ought to be more upfront about just what constitutes over-use. However when you take into account how diverse the plants that they own are (750Mhz, 850Mhz, 1Ghz?) + (DOCSIS 1.1, DOCSIS 2.0 DOCSIS 3.0) , and even the different nodes within systems (100 subs on node A, 17 subs on NODE B), you can understand why it is somewhat difficult to put in the TOS a black and white definition of overuse. Not to mention the fact that services and subscriber load on a system could change on a weekly or daily basis.

If they defined "overuse" according to the most heavily subscribed node, it would mean a lot of users who could be taking advantage of free bandwidth are being denied unused resources whereas a policy based on the least subscribed node would ultimately result in poor service for all. By addressing the issue of "overuse" on node by node, system by system basis they can ensure that their network resources are most efficiently use.

I don't think they intentionally try to make the definition of overuse a secret, it's just that it's based on so many dynamic factors it's almost impossible to put in black-and-white, but I'm sure it is easy for them to identify in the network management software which can dynamically evaluate all those factors.
Corydon
Cultivant son jardin
Premium Member
join:2008-02-18
Denver, CO

Corydon

Premium Member

said by TZi:

I too believe that Comcast ought to be more upfront about just what constitutes over-use. However when you take into account how diverse the plants that they own are (750Mhz, 850Mhz, 1Ghz?) + (DOCSIS 1.1, DOCSIS 2.0 DOCSIS 3.0) , and even the different nodes within systems (100 subs on node A, 17 subs on NODE B), you can understand why it is somewhat difficult to put in the TOS a black and white definition of overuse. Not to mention the fact that services and subscriber load on a system could change on a weekly or daily basis.
That's a good point, but Cox somehow manages to do it, and I'm sure they suffer from exactly the same issues that Comcast does.

A diverse plant actually is one of the areas where infrastructure upgrades does make sense. I have no doubt whatsoever that MSOs have network surveillance that monitor nodes for oversubscription, identify good candidates for upgrades, etc. I'm sure they do node splits and similar upgrades all the time as their business grows.

What I would like to see is a high cap on how much you can use the service in a month, one that would pretty much never catch 95% of their users (The MSOs routinely claim that only about 1% of users violate the invisible caps). Give people a way of tracking their usage on their website. Then selectively enforce the cap on those nodes where the heavy users are actually impacting others, while letting heavy users on other nodes that don't have the same impact skate by until they do.

Karl Bode
News Guy
join:2000-03-02

1 edit

Karl Bode

News Guy

quote:
That's a good point, but Cox somehow manages to do it, and I'm sure they suffer from exactly the same issues that Comcast does.
As an aside, Cox also uses the exact same packet forgery approach to throttle eDonkey traffic, but people generally ignored that when we pointed it out last year.
wierdo
join:2001-02-16
Miami, FL

1 edit

wierdo

Member

said by Karl Bode:
quote:
That's a good point, but Cox somehow manages to do it, and I'm sure they suffer from exactly the same issues that Comcast does.
As an aside, Cox also uses the exact same packet forgery approach to throttle eDonkey traffic, but people generally ignored that when we pointed it out last year.
Oh, they do it to BitTorrent in some markets (or at least were a couple of months ago)

Edited to add: As of a few minutes ago, they're not blocking outright anymore, they're just throttling the crap out of it. I have a server connected over GigE to Cox San Diego (among other ISPs) with an all Cox path to my cable modem in Tulsa. Using BitTorrent, the server can seed a given file to me in Tulsa at a mere 25.6KBps. Over HTTP, I can download the same file at my full cap.

It's wierdo, not weirdo. Yes, I know that's not the 'proper' spelling of the similar english language word.

funchords
Hello
MVM
join:2001-03-11
Yarmouth Port, MA

funchords to Karl Bode

MVM

to Karl Bode
said by Karl Bode:
quote:
That's a good point, but Cox somehow manages to do it, and I'm sure they suffer from exactly the same issues that Comcast does.
As an aside, Cox also uses the exact same packet forgery approach to throttle eDonkey traffic, but people generally ignored that when we pointed it out last year.
I'm 100% sure the same was happening with BitTorrent. My Cox partner in testing was having too many wireshark problems for me to be able to prove it, however. So the only evidence I had was anecdotal.

My confidence is based simply on the fact that Cox is using Sandvine and ED2K is an "also ran" as far as P2P protocols go. A USA cable MSO is not going to use Sandvine P2P policy enforcement without going after BitTorrent.

TZi
join:2001-07-05
Miami Beach, FL

TZi to Corydon

Member

to Corydon
said by Corydon:

What I would like to see is a high cap on how much you can use the service in a month, one that would pretty much never catch 95% of their users (The MSOs routinely claim that only about 1% of users violate the invisible caps).
The reason these caps are "invisible" is because it would ultimately mean issuing a cap on a node-by-node basis which would be ridiculous considering most users don't even know what a bandwidth cap is. I don't even really think that Comcast even has a number that constitutes abuse, they simply identify users on a node which are consuming resources to the point that it is causing performance degradation and dealing with them accordingly.

If they told you don't transfer over XYZ amount a month then 100 subscribers signed on next month, that number would no longer be relevant or effective. So too, if 100 users cancelled next month and you were forced to adhere to XYZ that would mean loads of network capacity go unused which could be used to satisfy you as a subscriber.

I don't think comcast intends to hide the "invisible" cap from us, it's just that the number is ultimately dynamic.

As for a webpage that details your total transfer, there are several programs for windows (in fact windows itself) and some routers that can keep track of your total usage. While it is easy for satellite providers who have one central NOC to provide such usage statistics, it would be a nightmare for comcast who is constantly acquiring older systems, migrating recently acquired systems and building out new systems to aggregate all this data onto a webpage.

funchords
Hello
MVM
join:2001-03-11
Yarmouth Port, MA

funchords

MVM

Our uplinks are too small to blame any one or any few users on a performance degradation. It's like trying to blame a particular rainstorm or a particular sunny day on El Nino or Global Warming.

That said, in the past, I think they have acted reactively -- in just the way that you described. They either looked at nodes with high utilization or looked at nodes with a high number of complaints. Then they made "the warning call."

But within the past year or so, there has been a stronger campaign and it seems to me that they may have made "the warning call" to certain users proactively -- perhaps where they were about to add a 16Mbps tier or where they were about to launch Comcast Digital Voice.
said by TZi:

As for a webpage that details your total transfer, there are several programs for windows (in fact windows itself) and some routers that can keep track of your total usage. While it is easy for satellite providers who have one central NOC to provide such usage statistics, it would be a nightmare for comcast who is constantly acquiring older systems, migrating recently acquired systems and building out new systems to aggregate all this data onto a webpage.
Assuming that you are right, then fair enough. However after "the warning call," the user should be able to get some periodic self-monitoring feedback from the provider. Perhaps someone can manually grab the data each week and send it via e-mail to those under "the warning." This way, a user at risk can understand whether the high usage is something nefarious (like a virus or a leeching neighborhood) or accidental (a lousy remote-backup configuration, or someone watching an HD-quality video stream 24/7 and merely turning off the monitor before going to school).

Currently, the only feedback they get is whether or not you have Internet service 30 days after that warning call.