| reply to DataRiker
Re: They make no sense. said by DataRiker:Well, there are ways to emulate http and https traffic And given the track record in this cat and mouse game, expect the p2p community to be 2 steps ahead. It shouldn't matter as long as they prioritize traffic based on how much bandwidth it is trying to use. If a certain type of traffic is peaking at 1.5MBps, it could be throttled during times of network saturation. Your VoIP, video chatting, and gaming (real time stuff) isn't going to max out your connection and hit really fast speeds. These things should be able to work uninterrupted. The stuff spiking are going to be huge web downloads, torrents, other download methods, hd video streaming, etc. Things that are not real time. They can be throttled down by half or more and the only effect is a longer buffer time and longer download time. It's possible and it's the way they need to do it. |
|
Reviews:
 ·RoadRunner Cable
4 edits | Your describing a throttle by volume, which only slows p2p transfers making them last longer (IMO making the problem worse in the long term)
This new system will have to deal with the following situation:
User A is using p2p with some type of https transfer
User B is trying to pay his credit card bill online
How will it judge which packet is real https and which is not?
I doubt this will have any better results than sandive.
If they drive p2p to this expect millions of upset users who can't get a secure web page to work. |
|
Lazlow
join:2006-08-07
Saint Louis, MO | reply to insomniac84
insomniac84
Why should one type of traffic receive higher priority than others? I do not VOIP and my neighbor does (we pay the same amount for internet). Why should his data get a higher priority than mine? Protocol agnostic is the only fair way to go. First come, first serve. |
|
patcat88
join:2002-04-05
Jamaica, NY
kudos:1 | reply to DataRiker
said by DataRiker:Your describing a throttle by volume, which only slows p2p transfers making them last longer (IMO making the problem worse in the long term) This new system will have to deal with the following situation: User A is using p2p with some type of https transfer User B is trying to pay his credit card bill online How will it judge which packet is real https and which is not? I doubt this will have any better results than sandive. If they drive p2p to this expect millions of upset users who can't get a secure web page to work. Rolling 7 day window of GB usage. Top 10 percentile on the node get lowest priority until they fall out of top 10 percentile. If node is never saturated, then there are no problems with being lowest priority. |
|
Lazlow
join:2006-08-07
Saint Louis, MO | patcat88
That unfairly penalizes those who do the majority of their downloading during non peak hours(which does not cause congestion or cost the ISP anything extra). |
|
patcat88
join:2002-04-05
Jamaica, NY
kudos:1 | Whats downloading? 50% of your limit off peak, or 100% of your limit? Rolling window usage to assign priority levels is fair. If your node is not saturated, you have no reason to care.
What about the person who wants to torrent off peak, but not 24/7? are they to be placed in the same class as the person who torrents off peak 12/7/365?
Rolling window with percentile will separate the HTTS bill payer from the P2Per, without content discrimination.
The less you use, the higher your peak speed will be, and the more bursty it will be. If you want a circuit, your going to get treated like a circuit.
If you don't have a rolling window (much better than a powerboost bucket), how do you separate the user which will release the channel when they get their "task" done (download a 700mb movie via HTTP), vs a user that will NEVER release the channel and can never finish their "task" (a heavy P2P downloader, note most P2Pers eventually finish their downloads, their uploads are unstopable and unfinishable, but some P2Pers have a hoarding mentality and select to download 100s or 1000s of files and everything that shows up in the search, they never keep or watch or use everything they download). |
|
Lazlow
join:2006-08-07
Saint Louis, MO | The ONLY time that downloading causes congestion is during peak hours. I do 99% of my downloading between 11pm and 8am (off peak here). This causes no congestion and does not cost the ISP anything extra. However I do like to browse (here and other places). Under your formula my browsing would be "throttled" because of my off peak downloading(which is causing a problem for no one). This is why a monthly or even daily download limit does not work. What needs to be done is to only count the usage during peak congestion. I can easily download(off peak ) 500GB+/month without contributing to the congestion while another person can download less than 150GB/month(1hr/night during peak congestion) and cause lots of congestion.
A pure proticol agnostic throttle during peak hours(only) is (in the short term) a good answer(which is what Comcast is using right now). The problem with this system is that there is the risk the ISPs will use it (long term) as an excuse not to upgrade their systems. |
|
1 edit | reply to DataRiker
said by DataRiker:This new system will have to deal with the following situation: User A is using p2p with some type of https transfer User B is trying to pay his credit card bill online How will it judge which packet is real https and which is not? That https p2p is going to make many connections to different hosts and not just one connection. And it will try to transfer much more data. Depending on how good you detect it, it might get a few seconds of max speed, then you start to throttle it back. |
|
2 edits | reply to Lazlow
said by Lazlow:insomniac84 Why should one type of traffic receive higher priority than others? I do not VOIP and my neighbor does (we pay the same amount for internet). Why should his data get a higher priority than mine? Protocol agnostic is the only fair way to go. First come, first serve. Well it can be protocol agnostic. I am saying look at sustained rates for specific transfers to specific hosts. VoIp isn't going to try to max out a 15mps connection, while bittorrent or other downloading will. You can leave lower bandwidth stuff alone since it's not a problem. Just throttle stuff that tries to max out the connection back enough to free up the congestion. It's way better to have a torrent drop back a few hundred kbps than it is to screw with stuff that isn't trying to use all possible bandwidth. There is no point in trying to throttle a 50kbps transfer, if another transfer is going 14.95mbps. Just throttle the faster transfer. And we are probably only talking between like 5-7pm, max. If at all.
Just throttle the specific traffic causing the problem, leave everything else alone. |
|
Lazlow
join:2006-08-07
Saint Louis, MO | insomniac84
What you have just described is what Comcast is doing now.
The time frame will also shift from market to market. Here 7pm-10pm will (usually) be more congested than 5-7pm. |
|
