Re: Faster... in http://www.dslreports.com/forum/r21373408 en Wed, 10 Feb 2010 02:24:37 EDT Wed, 10 Feb 2010 02:24:37 EDT Re: Faster... http://www.dslreports.com/forum/remark,21381618 anon : "You look up YOUR IPN (at whois.arin.net) and that tells you all the others who are serviced by your IPN Block. You then look up what other IPN Blocks your ISP has. In my case, I am in block NETBLK-OOL-4BLK and altering that 4 to 1-9 shows the other blocks (only blocks 3-6 exist). ..."

This is doable, but it's not as good as getting a network map from the ISP for a number of reasons.
- Mapping IP's to ISPs (i.e. ASNs) using public sources is only approximate (like your example of looking at similar netblock names). If the ISP provides a precise map of their IP prefixes, it is more accurate. And if it's easily queries through a standardized service (assuming that ISPs support it, etc.), it's easier to implement.
- The ISP knows their internal structure and resources/costs, so P4P guidance can optimize internal data flow. For example, knowing who is where inside Comcast's network allowed us to move a lot of traffic not only within Comcast but within the same hub or metro area, avoiding consuming long distance links.
- The ISP knows their available external links. For example, P4P allows an ISP to drive traffic to adjacent ISPs (e.g. those with direct/peering links) avoiding general internet traffic (which is generally slower and costs more).

There is also a subtle limitation to the "p2p like bittorrent will already try to use the faster peers over the slower ones" is that it only does so within the population of peers that are known to the peer and have sent the specific peer data. Early in a swarm's life a given peer only knows about a very small subset of the swarm, and has only tested actual throughput with those peers for an even smaller subset. So if you want to find the "best" peer out of a swarm of 100,000 peers, it could take a very long time to even get the complete list (from tracker announces, peer exchange, etc.). Then from that list, you would need to actually exchange data with one randomly selected peer a second (the generic BitTorrent 'optimistic unchoke' approach), it would take 100,000 seconds to have a 50% chance of finding the "best" peer. So, for very large files in large, the time to find the optimal peer would be much longer than the total download time.]]> http://www.dslreports.com/forum/remark,21381618 Wed, 05 Nov 2008 16:20:32 EDT Re: Faster... http://www.dslreports.com/forum/remark,21380777 RARPSL :
said by funchords See Profile :

Easy on the caps, chief. I'm happy to be wrong, just don't punish me for it.

There are a few technical issues with your explanation (around organization and server load), but you're not far from the stuff that we're talking about. Please look at the IETF p2pi and tana lists, catch up, and jump in if you can lend a hand. P2P experts are really welcome and needed.

But the point is that the P2P networks aren't doing this already. So it does take the 15-30 minutes that I mentioned.

Along side the P4P services ideas mentioned, there is an Azureus plug-in called Ono that leverages CDN locations.

Finally, we don't want to lose any rare sources or files, nor do we want to screen out remote peers, so any method used has to have a bias toward closer peers while not adversely impacting the performance for anyone. It may be that only P2P content that is very popular should use this at all.
OOPS. Sorry about the use of capital letters (Only the Netblocks should have been in caps - I should have used bold/underline/italic for the others).

I know that it is not being done now but I was just pointing out that it can be done in the future by adding support to the client. My method puts the support into the client without needing any help from an outside P4P server (only the standard Netblock WhoIs lookup).

Now to the technical details. At the current time, the clients have a user designated number of session slots devoted to handling a torrent and a list of peers to who can supply pieces of the torrent. There are routines to select which of these peers gets assigned one of the session slots. What method is used to make this selection is not part of the P2P Protocol and thus is probably different in different clients. Being "Greedy" by selecting Seeding (have 100% of the pieces) peers as opposed to Leaches (only have some of the pieces) gets you the torrent faster (assuming that the Seed and the Leach feed you at the same rate) since you do not need to devote time/effort to sending pieces. If are talking to a Leach who does not have any pieces you need and does not need any you have then the selection routine might drop the session to reuse the slot to select a Leach with pieces you need or even better a Seed. The selection routine might look for and contact Leaches that have needed pieces in the absence of Seeds (which always will have the pieces). There are a number of strategies that can be implemented. My suggestion just adds to the mix by favoring those peers in my Netblock followed by those in other Netblocks belonging to my ISP before going to peers in Netblocks that belong to other ISPs. This way I am playing "Nice Guy" by keeping my traffic local (ie: On my ISP's Network).

As to your "... rare sources or files, nor do we want to screen out remote peers ..." issue, if I have not filled my session limit with peers on my ISP's network, those rare and remote peers will still get selected to fill the quota (and the local peers will be rejected in exchange for remotes once the locals have no more pieces that are needed but remotes still do). In fact there can be special case code that allocates a portion of the slots to going after rare pieces by including those peers that have them. There is already such code for when you are already talking to a peer (you choose which piece you ask a peer to send you and thus can get the rarest pieces as soon as they become available). ]]> http://www.dslreports.com/forum/remark,21380777 Wed, 05 Nov 2008 13:20:07 EDT Re: Faster... http://www.dslreports.com/forum/remark,21375209 funchords :
said by RARPSL See Profile :

said by funchords See Profile :

said by dnoyeB See Profile :bittorrent will already try to use the faster peers over the slower ones. ... this means you will be using in-network peers ... So its already doing this.
Right, and the P4P guys at Pando and Verizon recognize this. However, finding those faster peers takes about 15-20 minutes.
WRONG. It takes only a few seconds. You look up YOUR IPN (at whois.arin.net) and that tells you all the others who are serviced by your IPN Block.
Easy on the caps, chief. I'm happy to be wrong, just don't punish me for it.

There are a few technical issues with your explanation (around organization and server load), but you're not far from the stuff that we're talking about. Please look at the IETF p2pi and tana lists, catch up, and jump in if you can lend a hand. P2P experts are really welcome and needed.

But the point is that the P2P networks aren't doing this already. So it does take the 15-30 minutes that I mentioned.

Along side the P4P services ideas mentioned, there is an Azureus plug-in called Ono that leverages CDN locations.

Finally, we don't want to lose any rare sources or files, nor do we want to screen out remote peers, so any method used has to have a bias toward closer peers while not adversely impacting the performance for anyone. It may be that only P2P content that is very popular should use this at all.
--
Robb Topolski -= funchords.com =- Hillsboro, Oregon
More features, more fun, Join BroadbandReports.com, it's free...
]]> http://www.dslreports.com/forum/remark,21375209 Tue, 04 Nov 2008 15:16:24 EDT Re: Faster... http://www.dslreports.com/forum/remark,21374498 RARPSL :
said by funchords See Profile :

said by dnoyeB See Profile :

I couldn't agree more. I was under the impression that p2p like bittorrent will already try to use the faster peers over the slower ones. Naturally, this means you will be using in-network peers over out-of-network peers. So its already doing this.
Right, and the P4P guys at Pando and Verizon recognize this. However, finding those faster peers takes about 15-20 minutes.
WRONG. It takes only a few seconds. You look up YOUR IPN (at whois.arin.net) and that tells you all the others who are serviced by your IPN Block. You then look up what other IPN Blocks your ISP has. In my case, I am in block NETBLK-OOL-4BLK and altering that 4 to 1-9 shows the other blocks (only blocks 3-6 exist).

At that point, you can prefer those peers who are being serviced by your ISP without any need for any special P4P server. All that is needed is to update the BitTorrent software to acquire this NETBLOCK info and use it. As the clients are updated, their efficiency goes up. Also this information allows the client who is looking for peers to connect with to first try the "local" peers before needing to go through a backbone to connect to the peer.]]> http://www.dslreports.com/forum/remark,21374498 Tue, 04 Nov 2008 12:52:51 EDT Re: Faster... http://www.dslreports.com/forum/remark,21373996 funchords :
said by dnoyeB See Profile :

I couldn't agree more. I was under the impression that p2p like bittorrent will already try to use the faster peers over the slower ones. Naturally, this means you will be using in-network peers over out-of-network peers. So its already doing this.
Right, and the P4P guys at Pando and Verizon recognize this. However, finding those faster peers takes about 15-20 minutes. The difference that P4P might make is ultimately going to be in those first 15-20 minutes.
--
Robb Topolski -= funchords.com =- Hillsboro, Oregon
More features, more fun, Join BroadbandReports.com, it's free...
]]> http://www.dslreports.com/forum/remark,21373996 Tue, 04 Nov 2008 11:17:58 EDT Re: Faster... http://www.dslreports.com/forum/remark,21373854 funchords :
said by jlivingood See Profile :

No one at Comcast has suggested charging for this -- only Karl has done some here.
Karl got it right, he merely failed to recognize that Pando already paid for this by providing Comcast with political cover during the FCC investigation. Thanks to Pando, Comcast was able to cast the impression that it was working with the P2P community. Pando wanted something, too. Trying to woo the NBC Direct deal, Pando's CEO hoped that their P4P deal with Comcast would ensure success.

Said Saul Hansell in the New York Times, "Robert Levitan, the chief executive of Pando, had told me that he hoped Comcast might program its network to give preference to applications like the one his company makes."

It was a pretty bogus deal at the time, but it did lead to some actual outreach and eventually I hope it will yield fruit down the road. I don't know the quote, something like "it has to get bad before it gets good." I think that's true and maybe that's what we were seeing back then.
--
Robb Topolski -= funchords.com =- Hillsboro, Oregon
More features, more fun, Join BroadbandReports.com, it's free...
]]> http://www.dslreports.com/forum/remark,21373854 Tue, 04 Nov 2008 10:55:17 EDT Re: Faster... http://www.dslreports.com/forum/remark,21373565 dnoyeB : I couldn't agree more. I was under the impression that p2p like bittorrent will already try to use the faster peers over the slower ones. Naturally, this means you will be using in-network peers over out-of-network peers. So its already doing this.

What Comcast proposes is to prioritize in-network peers even if they are not faster. That is the only difference I can see. But I can't see any reason why they would not be faster already!?

On the legality front I have to disagree with you. There is plenty of legal content on p2p. I p2p 1 illegal song years ago when it[p2p] first came out because the song is not available due to disputes. Since then I have p2p 1000s of gigabytes of legal content.]]> http://www.dslreports.com/forum/remark,21373565 Tue, 04 Nov 2008 10:02:40 EDT Re: Faster... http://www.dslreports.com/forum/remark,21373408 NOZIREV : i wonder if you dont use this new p4p if you will be throttled ???
--
"Citius, Altius, Fortius" [Faster, Higher, Stronger]]]> http://www.dslreports.com/forum/remark,21373408 Tue, 04 Nov 2008 09:36:58 EDT Re: Faster... http://www.dslreports.com/forum/remark,21373409 jlivingood :
said by swhitney2003 See Profile :

Currently I can download torrents maxing out my download speed. Why would I care about p4p? I'm not going to be downloading any faster than the speed I pay for. And a potential, additional charge for customers to attain this services of "prioritized p2p"... ridiculous. My speeds are fast enough, let alone paying for some shenanigans just so traffic stays in Comcast's network. Sounds like a win-win for the ISP here, get paid and use less inter-isp data.

Will this take off in the real world? Applied to torrenting? Or is it going to use proprietary software (ads infested too?)? Don't get me wrong, this idea is wonderful. It just has to sell to the public, who are already happy with torrents. #1 reason why it might fail... no 'illegal' content. Majority of p2p is illegal, so there won't be much left for P4P. Can this new technology really stand up and become a player in today's world?
No one at Comcast has suggested charging for this -- only Karl has done so here.

As this work moves into the IETF, you might expect that a standardized iTracker function was available with a well known name on a well known port, for any application to access on an open basis. As noted in the draft, the more clients making use of it, the better. So ISPs actually have a motivation to maximize use of the tracker rather than to constrain it via pricing mechanisms, proprietary APIs, etc.
--
JL
Comcast]]> http://www.dslreports.com/forum/remark,21373409 Tue, 04 Nov 2008 09:36:58 EDT Faster... http://www.dslreports.com/forum/remark,21373371 swhitney2003 : Currently I can download torrents maxing out my download speed. Why would I care about p4p? I'm not going to be downloading any faster than the speed I pay for. And a potential, additional charge for customers to attain this services of "prioritized p2p"... ridiculous. My speeds are fast enough, let alone paying for some shenanigans just so traffic stays in Comcast's network. Sounds like a win-win for the ISP here, get paid and use less inter-isp data.

Will this take off in the real world? Applied to torrenting? Or is it going to use proprietary software (ads infested too?)? Don't get me wrong, this idea is wonderful. It just has to sell to the public, who are already happy with torrents. #1 reason why it might fail... no 'illegal' content. Majority of p2p is illegal, so there won't be much left for P4P. Can this new technology really stand up and become a player in today's world?]]> http://www.dslreports.com/forum/remark,21373371 Tue, 04 Nov 2008 09:27:33 EDT