Building a Better P2P P4P coalition offers more efficient delivery... Thursday Aug 21 2008 12:41 EDT The P4P Research Group (pdf), a coalition of most major ISPs, researchers and Pando networks, is working on a more efficient P2P protocol that saves transit time by only serving file parts from local peers to reduce hops. Pando and the new coalition believe they can speed up P2P transfers by as much as 235% across US cable networks and up to 898% across international broadband networks. In Verizon tests, Pando increased the percentage of data routed internally across their networks from 2.2% to 43.4%, which they claim reduced inter-ISP data transfers by an average of 34% (up to 43.8 % in the US and 75.1% internationally). The project should get renewed attention this week as researchers from the University of Washington and Yale University plan on releasing additional data from project tests at a presentation late today in Seattle. While the possibilities of the technology are promising, there's a lot of questions concerning how this would be implemented. Would ISP partners (AT&T, Verizon and Comcast are involved in testing) charge customers more for prioritized P2P? Will the client source code be published? Will the system come with anti-piracy provisions and if so, will it create an ISP gatekeeper situation to wall off "non-sanctioned" P2P content? |
|
Must be open source!Must be open source! Or I will not trust it. | |
| | swhx7 Premium Member join:2006-07-23 Elbonia |
swhx7
Premium Member
2008-Aug-21 12:46 pm
Re: Must be open source!said by AstroBoy:Must be open source! Or I will not trust it. That would be a necessary condition for me too. But even if it's open source, if it's run by the ISP, it's still untrustworthy in my view, and will be closely monitored. The more important question is about: quote: Will the system come with anti-piracy provisions and if so, will it create an ISP gatekeeper situation to wall off "non-sanctioned" P2P content?
That's probably the goal: make the p4p easy and fast, but it will have only the selection approved by the ISP and the copyright holders, and maybe laden with DRM, fees and whatnot. Then they'll try to demonize regular p2p, implying that anything outside the approved application must be piracy. Then throttle, ban or otherwise interfere with regular p2p, based on a claim that p4p should be able to do everything the user wants. | |
| | | NerdtalkerWorking Hard, Or Hardly Working? MVM join:2003-02-18 San Jose, CA 1 edit |
Re: Must be open source!I've heard about this P4P conglomerate before, and although I'm not sure whether it's OSS, they're very very reputable. | |
| | | | funchordsHello MVM join:2001-03-11 Yarmouth Port, MA |
Re: Must be open source!I think the process that they're taking is a bit "Rube Goldberg"-ish. I've talked to Laird Popkin at Pando about Open Source and he assures me that it will be Open Source. However, the ISP's data (the input that drives some of this behavior) might be closed or obfuscated somehow (I really haven't looked into that). Some ISPs are sensitive about competitors using the data to learn their secret recipe.
I still haven't taken a position on it, but my fears have gone from "be afraid, be very afraid" (especially concerning the who's who that is in the DCIA) to one of "as long as users will have free choice of applications and protocols, then this is just another market entrant and it will live and die on its own merits."
I am put off with the test results of 235% and 898%, since the results are not described in terms that can be independently reproduced. But that's just the technician in me. | |
|
| | |
MrMoodyFree range slave Premium Member join:2002-09-03 Smithfield, NC |
MrMoody
Premium Member
2008-Aug-21 12:12 pm
Statisticsquote: they can speed up P2P transfers by as much as 235% across US cable networks and up to 898% across international broadband networks.
Is it just me, or do these sound like made up numbers? | |
| | |
Re: Statisticsquote: Is it just me, or do these sound like made up numbers?
Never trust round, even numbers. | |
| | | R4M0NBrazilian Soccer Ownz Joo join:2000-10-04 Glen Allen, VA 1 edit
1 recommendation |
R4M0N
Member
2008-Aug-21 12:26 pm
Re: Statisticssaid by Devorius:quote: Is it just me, or do these sound like made up numbers?
Never trust round, even numbers. I find that highly offensive! Sincerely, 80 | |
|
| |
to MrMoody
said by MrMoody:quote: they can speed up P2P transfers by as much as 235% across US cable networks and up to 898% across international broadband networks.
Is it just me, or do these sound like made up numbers? 898% of a few kbps, say 10Kbps, is only 89.8kbps. It is like saying that the murder rate increased 100% in a small town, even though they only had two murders because last year they only had one. Without meaningful numbers, like how many kbps the average file transfer is before and after, those "increased by" percentages are worthless, just like in the murder rate example. | |
| | swhx7 Premium Member join:2006-07-23 Elbonia |
to MrMoody
The promised speedup is based on preferentially getting files or pieces from within the ISP's network, and going out to other peers elsewhere as little as possible.
Of course that's the advantage for the ISPs, they don't have to pay the big-pipe providers for "internal" traffic, so to the extent they get their customers using this instead of regular p2p, they save on "external" traffic charges. | |
|
jjoshua Premium Member join:2001-06-01 Scotch Plains, NJ |
jjoshua
Premium Member
2008-Aug-21 12:44 pm
Sounds too easyI get the concept and it makes sense - get data from the closest node.
Why don't current P2P clients just use your network subnet to make an best guess as to which nodes are closer and give them priority? | |
| | swhx7 Premium Member join:2006-07-23 Elbonia |
swhx7
Premium Member
2008-Aug-21 1:10 pm
Re: Sounds too easyquote: Q: Why cannot P2P achieve the benefits of P4P by itself? A: In the current Internet, for P2P to explore peering flexibility to improve network and application efficiency, it will have to probe the network to reverse engineer information such as topology, and network status. This is however rather challenging in spite of significant progress in network measurement techniques.
Reading the pdf reveals it is more like a protocol and interface to enable p2p and other applications to get info from the ISP (or a "trusted third party") about network conditions, and cues about which routes to use. So p2p apps will need modification to use this, and can't do it alone because it relies on the ISP or special server ("i-tracker") to serve info on preferred routes. The client could be programmed to prefer local links, but p4p lets the network management people manage it by indicating which links are preferred for a combination of reasons, whether they're local or remote. | |
| | | Cabal Premium Member join:2007-01-21 |
Cabal
Premium Member
2008-Aug-21 3:37 pm
Re: Sounds too easyEven so, P2P apps could (and should) still make a significant difference by giving preference based on latency. It wouldn't be perfect by any means, but favoring a connection with 20 ms latency is probably much more likely to be closer/local than one with 200 ms latency (I would say "ping time", but that would be confused with something ICMP-specific). | |
|
| KearnstdSpace Elf Premium Member join:2002-01-22 Mullica Hill, NJ |
to jjoshua
i see this benefiting more of the legit P2P uses like the blizzard downloader. where companies could peer with say Comcast and have the seed itself with in the network meaning no 3rd party network uses needed. | |
|
|
simpleCouldnt we just do this as the following:
1. figure out your pubic IP 2. look up the ASN of your IP 3. look up the ASN of your potential peers 4. send chunks more often/faster to peers with same ASN as you
And save the hassle of all these lobby agencies? | |
| | | | jap Premium Member join:2003-08-10 038xx |
to patcat88
said by patcat88:Couldnt we just do this as the following: 1. figure out your pubic IP 2. look up the ASN of your IP 3. look up the ASN of your potential peers 4. send chunks more often/faster to peers with same ASN as you My thoughts exactly and it's probably what ISPs will through into the mix. I'm guessing the consortium is to share design techniques and get the networks to adopt it quickly. It's about edge costs anyway, not making things faster for us. Oh, I just checked and my pubis doesn't have an IP. How do I get one? | |
|
|
who is sharing?I'm on Comcast's network. The majority of peers I connect to when downloading don't even have Comcast. So that makes this all completely useless if the people I download aren't even on the same ISP. | |
| | |
Re: who is sharing?It's not about privileging peers that are on the same ISP, but rather those that are "close". My IP peers on the Ottawa Internet Exchange, so this thing would tell my p2p client to connect to a [p2p]peer whose ISP peers with mine. It would not tell my client to connect to a [p2p] peer that's connected to a PoP at the other end of the country, even if he's with the same ISP. Ugh, the ambiguity of the word "peer" kinda killed my message. | |
|
FLengineerCCNA, CEH, MCSA Premium Member join:2007-06-26 Deltona, FL |
P2P and IPv6If we could ever get IPv6 to take off then P2P would be able to do what P4P is supposed to do by only changing the clients instead of the entire protocol. | |
| RARPSL join:1999-12-08 Suffern, NY |
RARPSL
Member
2008-Aug-21 3:03 pm
There Must be Peers on my ISP's Network for this to workThis method ASSUMES that there are enough peers on my ISP's Network for me to not need to go to other ISP's networks. In my use of P2P this never happens or if I do have a local peer, it is the only one. Thus either I will not be able to get my data (if I am blocked from connecting to a non-ISP peer) or I will get it VERY slowly since I will only be being feed by the Local Peer.
I can see the BT clients being upgraded to look at the Network Topology by using the ASN (as another poster suggested) and preferring those peers just like they currently do for faster peers and those that deliver the pieces requested by going with "Roll Your Own" P4P type support. This can take the course of upgrading the P2P handshake to have the peer supply its ASN or by having the Client do the ASN probing itself. In either case once you know the ASN of the peers you are talking to, you can use an ASN match to assist the preference algorithm to prefer peers on your network to those on other networks IF all the other preference factors are equal for the two peers. Just preferring a local peer over a better non-local peer is not going to give the same performance as at present for the reasons given above. For this to work, you will need to go to all available peers and then tune your preferences to prefer the local ones as you run IF they are delivering the same or better than the non-local peer (remember that the local peer will be able to deliver faster than a non-local peer due to less latency, hops, and the size of the Internet Pipe). | |
| KylemaulLovin' My Firefox Premium Member join:2001-03-30 Puyallup, WA |
Kylemaul
Premium Member
2008-Aug-21 4:44 pm
Better for whom?This absolutely reeks of yet another effort by the corporations to undermine the current model of file sharing to serve their own ends. Will it be faster? Sure, even if only slightly. Will it be as anonymous as current file sharing applications? You can bet the answer will be a huge "NO". Most likely it will simply provide traceability to give those going after copyright infringement violations a leg to stand on in court. And for the mere possiblility that that scenario is highly likely, it adoption rate will likely be very slow. With the nature of current file sharing being what it is, this fact alone will mean that the selection of files available will not be what those who use these types of applications are after. I give this project a big FAIL from the outset. Just my prediction. | |
| GlaiceBrutal Video Vault Premium Member join:2002-10-01 North Babylon, NY |
Glaice
Premium Member
2008-Aug-21 5:00 pm
No greedy companies!The last thing is needing those scummy **AAs being involved in this! | |
|
| |
|
|