how-to block ads
|
Uniqs:
1498 |
Share Topic
 |
 |
|
|
 jlivingoodPremium,VIP
join:2007-10-28
Philadelphia, PA
kudos:1 | Slight clarification / correction Comcast says that sustained use of 70% of your up or downstream throughput triggers the BE state, at which point you'll find your connection at 50% speed for a period of fifteen minutes. The 50% of that is being misunderstood. At no point is traffic "throttled" to a specific speed. I think you are misreading the re-entry criteria to go back into PBE.
Before we even look at per-user consumption, the CMTS must have entered a Pre-Congestion State. Once this is the case, and a congested time may be approaching, user traffic may be marked as BE instead of PBE. You are correct that the upstream or downstream %s are 70% of the provisioned amount. So for someone on the 12Mbps/2Mbps tier, that would be 8.4Mbps downstream or 1.4Mbps upstream, over 15 mins.
The 50% figure does not mean that someone is throttled to 50% of their speed. It means that in order to have traffic marked as PBE again, from BE, that the upstream or downstream (whichever is managed) used over a 15 min period must fall below 50% of provisioned. Using the example 12/2 speed above, that is 6Mbps down or 1Mbps up (thus this varies by speed tier) over the measurement period.
Please feel free to refer to the FCC doc on this at »downloads.comcast.net/docs/Attac···ices.pdf.
Regards
Jason
--
JL
Comcast | |
|  | | Re: Slight clarification / correction Thanks, I realized I had that wrong and fixed it as I was talking to a Comcast spokesperson on the phone. | |
| | jlivingoodPremium,VIP
join:2007-10-28
Philadelphia, PA
kudos:1 | Re: Slight clarification / correction said by Karl Bode:Thanks, I realized I had that wrong and fixed it as I was talking to a Comcast spokesperson on the phone. Thanks. As a small nit, which I don't expect anyone to correct here or on other stories due to the good nature of the sound bite, the system is not accurately a 'throttling' system but is instead a 'packet prioritization' or 'QoS' system. If we were to implement a true throttling system, it would act in a very different way, by throttling down traffic to set speeds at certain times of day (a la systems in the UK) or under certain conditions (a la some wireless networks) or for certain application protocols. 
Jason
--
JL
Comcast | |
| | |  funchordsHelloPremium,MVM
join:2001-03-11
Yarmouth Port, MA
kudos:5 | Re: Slight clarification / correction said by jlivingood:As a small nit, which I don't expect anyone to correct here or on other stories due to the good nature of the sound bite, the system is not accurately a 'throttling' system but is instead a 'packet prioritization' or 'QoS' system. No, Jason, that's equally inaccurate, since you're dumping the whole of a customers traffic into it -- that's subscriber prioritization, maybe. To call it QoS or packet prioritization is like calling Comcast's network "fiber."
--
Robb Topolski -= funchords.com =- Hillsboro, Oregon -- KJ7RL
What you do at Christmas does not matter so much; What counts are the Christmas things you do all year through. | |
| | | | miscDude
join:2005-03-24
Hendersonville, NC | Re: Slight clarification / correction said by funchords:No, Jason, that's equally inaccurate, since you're dumping the whole of a customers traffic into it -- that's subscriber prioritization, maybe. To call it QoS or packet prioritization is like calling Comcast's network "fiber." With QoS and packet prioritization, my understand is that you can assign the criteria which to base it's priority on. In this case, Comcast is probably basing that priority off the IP or MAC address of the customer.
While most QoS services will base it's Quality of service based on the packet type or contents, (UDP vs. TCP vs etc), Because of the political needs to have a protocol agnostic way of managing their network, they had to choose a different way of differentiating the traffic.
I may be wrong, but from the sounds of it, it still is technically a QoS / Packet Prioritization system... even if it may use different criteria in differentiating those packets then what we are used too dealing with when talking about QoS.
| |
| | | | | funchordsHelloPremium,MVM
join:2001-03-11
Yarmouth Port, MA
kudos:5
2 edits | Re: Slight clarification / correction said by miscDude:While most QoS services will base it's Quality of service based on the packet type or contents, (UDP vs. TCP vs etc), Because of the political needs to have a protocol agnostic way of managing their network, they had to choose a different way of differentiating the traffic. No, they didn't have to do this at all. They simply can stop selling higher-speed tiers in areas not yet equipped to handle them.
But if they want to offer QoS, it needs to be done such that the network isn't making these value decisions. That was the political problem they had: they were judging what connection lived or died and doing so whether or not there was an actual bandwidth problem.
Yet they've always been in control of how much traffic is admitted on to the net. They sell tiers and are in complete control of whether or not they add a subscriber.
But beyond that, if they want to offer QoS, they can simply let the network be sensitive to DSCP or TOS markings placed there by the customers' applications. They can quota the fastest markings (to avoid abuse) and have no cap on the scavenger ones. With that, arguably unlimited broadband with the customer in control. Doing so would do no harm and would make the entire Internet more elastic (if the nation's largest ISPs started recognizing handling instruction markings appropriately, the worldwide tech industry would follow).
Instead, we've got this complicated proprietary rather stupid non-transparent sledgehammer system with thresholds so high one wonders whether it really does them any good at all. The best we can hope is that it does no harm to us.
--
Robb Topolski -= funchords.com =- Hillsboro, Oregon -- KJ7RL
What you do at Christmas does not matter so much; What counts are the Christmas things you do all year through. | |
| | | | | |  ThrowDemsOutIf you can't convince 'em, confuse 'emPremium
join:2002-03-03
Mullica Hill, NJ
kudos:4
3 edits | Re: Slight clarification / correction said by funchords:But beyond that, if they want to offer QoS, they can simply let the network be sensitive to DSCP or TOS markings placed there by the customers' applications. Yeah, that will work{/sarcasm}. So every bandwidth hog P2P user would be using a P2P app that set their app as the highest priority traffic based on TOS/DSCP settings.
They can quota the fastest markings (to avoid abuse) And thereby get right back in to the mess of throttling based on application &/or protocol.
Face it, you just want to have all ISPs provide unlimited bandwidth with no restrictions at all on the worst abusers of the system. It is the logical consequence of all your suggestions.
--
My BLOG .. .. Internet News .. .. My Web Page
Ask yourself one question: 'Do I feel lucky?' Well, do ya punk? | |
| | | | | | | funchordsHelloPremium,MVM
join:2001-03-11
Yarmouth Port, MA
kudos:5 | Re: Slight clarification / correction said by ThrowDemsOut:They can quota the fastest markings (to avoid abuse) And thereby get right back in to the mess of throttling based on application &/or protocol. How do you figure? Putting a quota on a DSCP means that it reverts to default behavior if the quota is passed. The ISP never has to look at application/protocol.
said by ThrowDemsOut:Face it, you just want to have all ISPs provide unlimited bandwidth with no restrictions at all on the worst abusers of the system. It is the logical consequence of all your suggestions. Face it, you don't pay attention to what I say. If you did, you would see:
•I have consistently called for abusers to be removed -- not coddled, not warned -- removed.
•I have consistently called for ISPs to disclose what they're doing so that if it's not working, that users can get support.
•I have consistently called for standards to be followed.
As for limits, I have said two things -- users need a fair meter and limits that don't grow stunt the growth of innovation.
--
Robb Topolski -= funchords.com =- Hillsboro, Oregon -- KJ7RL
What you do at Christmas does not matter so much; What counts are the Christmas things you do all year through. | |
| | | | | | | |  espaethDigital PlumberPremium,MVM
join:2001-04-21
Minneapolis, MN
kudos:2
 ·Clear Wireless
| Re: Slight clarification / correction said by funchords:How do you figure? Putting a quota on a DSCP means that it reverts to default behavior if the quota is passed. The ISP never has to look at application/protocol. Only the users that configure their applications to set DSCP values will benefit from such an implementation.
Even with per-connection quotas, it's a pretty safe assumption that a minority of end-users will have applications that will mark traffic. It's also a pretty safe bet that P2P apps will be disproportionately represented in that group. | |
| | | | | | | | | funchordsHelloPremium,MVM
join:2001-03-11
Yarmouth Port, MA
kudos:5 | Re: Slight clarification / correction said by espaeth:said by funchords:How do you figure? Putting a quota on a DSCP means that it reverts to default behavior if the quota is passed. The ISP never has to look at application/protocol. Only the users that configure their applications to set DSCP values will benefit from such an implementation. Even with per-connection quotas, it's a pretty safe assumption that a minority of end-users will have applications that will mark traffic. It's also a pretty safe bet that P2P apps will be disproportionately represented in that group. As I said -- if Comcast (or AT&T) were to do it, I think the entire world would follow.
Actually, I predict the VOIP and gaming apps would be the first to rush to take advantage of such a thing, while P2P file-sharing apps themselves would either opt for best-effort or less-than-best-effort (a lower-priority bulk class, if such were made available).
I'll say that again, I don't think there is any interest among the community of P2P developers to harm the network. They know that an efficient network works better for them.
Sure -- It's predictable that hacks and mods would come out that tried to exploit faster markings, which is why I think they should be quota'd by the ISPs (actually, I think Brutus the Enforcer should pay the snot a personal visit, but that's just me). An ISP can quota a VOIP/Gaming DSCP to 3x-5x the width of the loosest of the VOIP codecs and still avoid making cheating at all attractive to a P2P user.
--
Robb Topolski -= funchords.com =- Hillsboro, Oregon -- KJ7RL
What you do at Christmas does not matter so much; What counts are the Christmas things you do all year through. | |
|
| |
|
