Harvard Develops P2P Client that uses Bandwidth as “Currency”
Written by Ernesto on August 30, 2007Harvard researchers have teamed up with the Tribler team to work on a P2P client with BitTorrent support that uses bandwidth as a global currency. They released Tribler V4.1 yesterday.
Yes, Harvard, the richest University in the world recently started a new line of P2P research. They have an army of law professors to protect them, so unlike others, they must feel safe to do this controversial research in the land of the free and the home of the RIAA/MPAA.
The Harvard project is all about a fresh new approach. To be honest, have we seen a new trick since eMule and Bittorrent started? Things have clearly slowed down in the last years.
The Harvard researchers are currently working on one of hardest P2P problems, ensuring uploads. P2P dies or thrives depending on how much upload people donate. By introducing electronic “currency” for uploads they think they can make P2P HDTV Video on Demand possible. With the minor detail that we all have to switch to their client…
The latest version of Tribler enhances the standard tit-for-tat BitTorrent algorithms with something they call the give-to-get algorithm (PDF article). This new algorithm allows their users to benefit from a good ratio without using a central server like private BitTorrent trackers do.
Tribler users can still join every BitTorrent swarm and play the tit-for-tat game with old-school BitTorrent users. But, when they meet another Tribler peer they switch to give-to-get mode where the currency meter is running. This turns the Tribler network into a private Tracker network without the central server. This basically means, the more you share, the faster your downloads will go.
Every Tribler client keeps an eye on MByte counts of fellow peers. They gossip around about who is a leecher and who is a top dog, without the details of which Hollywood movie it was. The only information displayed about this in the GUI is a list in your profile of the “Top 10 Tribler Uploaders”. For the next version of Tribler they plan to turn that list of top dog uploaders into a decentralized trust system and enable users to correct typos and add tags to the content. In short, BitTorrent would go “2.0″.
But let’s first see if they can really handle network pollution and spam without a central server. It will be quite tricky to get such “Google PageRank” trust algorithms working in P2P.
Previously: ‘Heroes’ Star Recognizes Benefits of BitTorrent
Next: Jamendo: Download Thousands of Free and Legal Music Albums
63 Responses
This will never work. Tit-for-tat is perfect. Give-to-get won’t work because the top dogs would take all the upload for themselves, thus you get no swarm.
@Bram : No I guess this is the way to go to reduce excessive leeching on bit torrent.
the software crash when i Launch it on OsX. sounds quite pointless anyway.
@Karthik Kastury
When you get no swarm, you get no leeching also, I guess. :)
When I try to launch it on OSX it also crashes X_X. Oh, little note, Harvard isn’t the first to do this. TvTorrents uses your upload bandwith as currency. No ratio, but whenever you download you lose credits, when you upload you gain them. You have to have a certain number of credits to start and maintain a download forcing you to upload. It’s really neat. Also, upload a lot more than you download and you get faster speeds and can download more at once.
looks like brad cohen’s mad because HE didn’t develop bittorrent 2.0 hahaha
bittorrent, or at least utorrent, gives you better download speeds depending on your ratio IIRC so leachers don’t always get the fastest speeds. this way does seem more efficient though. let’s see how it plays out
With all respect to Bram: come on, how much leeching goes on all the damn time out there with tit-for-tat today? Private trackers tend to give a world of better experience than the constant plague of hit and runners … which is human nature after all once they’ve finished their downloads.
Ratio enforcement has been an issue for ages now; including hacked clients which try to screw the trackers. Decentralised trust ranking and record seems like a great idea if only it can be implemented well. Or at least it seems like a candidate.
One problem I see is the IP drift all cable modem users are subject to. And it’s the ocean of asymmetrical pipes through cable modems which gives rise to ratio issues in the first place. Seriously: most of us have to spend 5-10x as long as a seeder than a leech to hit 1:1. Do all of us do it? Hell no. How much that’s hurting torrents overall is an open question.
@Brian
No, I’m not mad. :) I’m somewhat amused by this whole idea, too bad it’ll fail miserably. Let’s for the sake of argument assume that I have a gigabit upstream, which I actualy do, not that you have to believe me. How many people with a single megabit upstream does it take to compete with me? How can anyone beat my dozen seedboxes? People share because they care, not because of ratio enforcement.
And, no. uTorrent and Mainline do not give you any faster downloads if you have a high ratio. If you don’t believe me, just check out network statistics and snoop around wireshark to see what is actualy going on. :)
@John M
Hit and runners are not the plague. The very idea behind BT is sharing the wealth. If someone cannot seed back what they’ve downloaded, is that the reason to punish them, and to make it impossible for them to download? Have you ever tried to seed back a torrent content that has 250 seeders on each leecher? Do you realy belive that the mathematical model of sharing requires to have so many seeders, or is it because of the private BT trackers and their bussiness of selling upload credit (seling users their own bandwidth)?
I wish the Tribler team the best of luck. They’re going to need it. :)
Over and out.
Though I understand how run-away leechers are a problem, just like John M. pointed out, some of us have suffer from pathetic upload rates and less-than-decent downloads, and even though I’d love to give a fair 1.1 (better 2) ratio, it’d be really difficult unless I left my computer running twenty four hours a day (which I can’t do). My download ranges from 50 - 57kB/s and my (suggested) upload is 11kB/s, I’d have to upload for fives times my download ETA. When/if I download things over 1Gb, I’m already talking about many long hours of uploading, and I’m sure many people suffer from said uneven download/upload, it’d be slightly discriminatory to those without grand connections.
It has also become a problem, personally, that for some reason, uTorrent is slowing down my download (for normal internet browsing) when I leave it to just seed (which I try to do as often as possible). I’ve already modified all the settings as to not cap my upload and supposedly allow a “free” browsing experience while uploading but it’s slowing it down significantly (pages loading anywhere from 4kB/s to 15kB/s where as before it’d allow me to reach 40kB/s, sometimes reaching 50kB/s).
In anycase, unless they can manage to conjure up a way so that the protocol understands these handicaps for some users, it won’t be all too fair.
The only real problem with BitTorrent is when the swarm dies, the file is no longer available. Project Osprey by ibiblio has already solved this.
http://osprey.ibiblio.org/
Bram Cohen is a fake!
leave it up to Harvard to make something into Currency, big shocker there.
this project sounds good on paper, but time will shed more light on this idea.
Well, I wish I could look into this but, being on Rogers SUCKS! Id have to seed like for 10 days to go 1:1 on some items.. Ive sort of prepared for the day Rogers would catch up and im using my good ratio which is decreasing due to not being able to upload) I think issue #1 should be to fix\change the ecryption to stay 1 step ahead of the nasty ISPs… Bram get this an ISP which advertised Bit Torrent at one time and now = excessive throttle on everything.
As far as I know, P2P Torrents are born out of people’s spirit of sharing. Sharing means giving something without expecting something in return. For me, law and technology can never understand that. Those people trying to make money out of free downloads are contradicting their own words. “The more you SHARE, the faster your downloads will go,” eh?
This would never work.
Every Internet service provider on Earth would unite and ban the file sharing that supports using upload bandwidth as currency.
As soon as your everyday home broadband Internet connection starts seeding, trying to compete with those whom have much greater upload bandwidth, it would end.
Companies like Comcast would not put up with their subscribers trying to complete with ISPs in western Europe or Japan that on average have 10 times the bandwidth up and down, for less cost.
Makes it easy for the top dogs to get busted and sources lost in an instant.
Well, eMule already works like that!
Upload more, you download faster.
With all the problems with our currency-based economy, it’s hard for me to imagine why we would want to emulate it.
Right off the top of my head, I think such a model emphasizes the “currency” over actual sharing. File sharing optimization focuses on increasing general sharing speeds, not on increasing available bandwidth. This is a crucial distinction that will cause this project to tank.
Since optimization occurs as an increase in _actual_ transfer speeds, and not in “potential” speeds (as in the currency model), then naturally the algorithm should address the actual transfers and not the potential ones (I’m using “potential transfers” to refer to bandwidth, which may or may not be used for actual transfers). As Bram has pointed out, the latter model will have wasted bandwidth (e.g. in the “250 seeders to 1 leecher” scenario), since it addresses potential transfers, a.k.a. “bandwidth”, and therefore its subjects may or may not actually materialize as real transfers; the ones that don’t result in wasted bandwidth. Wasted bandwidth = 250 seeders sitting around, wasting resources on both ends, just to keep a good ratio.
Join a private tracker and run a torrent that has a 250:1 s/l ratio and see how long it takes to keep a good d/u ratio. The ratio is not important at all, since leeching without seeding will hardly negatively affect the swarm at all, yet you will still be punished with a poor ratio. Since the focus is on the node’s ratio rather than on the swarm’s sharing efficiency, it results in such a very sub-optimal situation. A good ratio doesn’t necessarily imply efficient file sharing, which is why Bram so vehemently opposes private trackers that focus on ratios.
Read up on game theory and the Prisoner’s Dilemma (which is directly translatable to file sharing). In game theory lingo, the “currency” model focuses on winning matches, not on winning points. For the Prisoner’s Dilemma, tit-for-tat has so far been the best overall strategy. That is a mathematical fact.
this won’t work well at all.
Give-to-get sucks
For Diversity and Creation , I think it’s a good idea.
Good luck to them, because they’ll need some. They’re definitely a setp ahead of Kazaa when using ratings - as Kazaa used local machine to store ratings (easily manipulated) and BT2.0 will use decentralized ratings. LOL. But the problem doesn’t end here. Suppose a hacker creates a BitTyranny2.0 program which floods the swarm with artificial clients that *cheatingly* gossip positive things about you especially when you’re leeching. That would be enough to exploit the ratings system used in BT2.0
So sorry Harvard, but as long as any “currency” is used there will always be counterfeiters.
@Bram
Very true criticism of the worst side to private trackers. I even saw a pretty decent one I was in once move to enforce an “everyone
@Bram
Very true criticism of the worst side to private trackers. I even saw a pretty decent one I was in once move to enforce an “everyone less than 1.0 is OUT!” rule … presumably to their own destruction. There are certainly a lot of tracker admins who also happen to be morons out there! The key is to find good ones.
Demonoid is probably the best example of a massive private tracker which keeps functioning AND gives me room to upload to a good ratio without the tyranny of “rare leecher” swarms. I always make a point of prioritising the swarms where my (slow) seed is actually of significance. At Demonoid that means leaving popular ones at 0.1 ratio while sticking with smaller ones well above 5.
@Anonymous #9
I have a one word answer for your problem: Autospeed. Azureus isn’t everyone’s cup of tea, but that function when well tuned is your best friend. Basically it detects when your pipe is getting clogged and throttles the torrents back so browsing etc. still works. Frankly essential for any shared home cable, and frustrating so many good clients like µtorrent don’t feature it.
I have a coder bud who’s building this feature into Transmission (a svelte native Mac client) to save me the bother of Azureus’ Java Achilles Heel. So far it’s testing quite nicely.
@Varun S
Counterfeits indeed. A good point. But notice how counterfeiting still doesn’t force the real world off hard cash? It’s kind of like Google Bombing … sometimes it’s a problem but overall Google and the world’s advanced economies work well. Comes down to balance I expect.
This sounds like bittorrent gone capitalistic, which is the last thing anyone needs. It creates a system wherein the haves (broadband users) get everything while the have-nots get screwed with no recourse.
Any advantage one might think this provides is outdone by the fact that you have to switch to their client. The “problems” it solves aren’t worth the effort.
A message from the Tribler team…
Nice to see that our idea has gotten a bit of attention. As it will be impossible to convince people. Please just wait a month or two and we hope to deliver better software.
We are improving a P2P client which has a build in understanding of people and Internet access asymmetry. Some private trackers have done a good job with ratio enforcement, other not. It is important to understand that is very difficult economic problem is at the hart of such systems. Most private trackers suffer from the problem which economists call the Babysitters problem… Details : http://arxiv.org/abs/0705.4110
(PDF, not a PNG)
Fortunatly not everyone needs to seed a lot. There are people who are willing to seed for a long time, just to be Number One in a highscore ranking of uploaders. So 1 guy gets social credits and gives away bandwidth for free to others.
I think it is important to facilitate the human side in P2P (more 2.0 buzz here)
Autospeed is a very good remark. I think an important improvement for todays client is more network-awareness. This is one of our research topics :
https://www.tribler.org/TriblerResearchSubjects
Johan.
what about those of us who WOULD contribute were it not for the fact that comcast clearly is blocking seeding?
I’ve had an idea for the second thing they mentioned. I hope they flesh it out as much as I have in my dreams, because I’d never have good enough programming skills.
whats new about this?
It’s always been like that on private servers, except they making it decentralized…
unless… this is the way they trying to get the user data since torrentspy refused to giveout its database
nice way to lurk in unsuspecting users
If the company’s will pay you some amount of “currency” to use your bandwidth to help share paid media won’t that violate some company’s TOS’s on bandwidth usage?
With all due respect, to Bram Cohen, isnt that an extreme example? I dont pretend to know much about the tech side of bittorrent, but dont the vast majority of users have roughly the same broadband speed?
@Bram:
As long as these guys aren’t going to use this system in an ultra-capitalistic way (and even then, they could get people to pay real money for extra download speed), I could see this thing working.
The share ratio measures not only a seeder’s potential for upstream, but also their altruistic nature. If you have a gigabit upstream (I’m jealous) and a tendency to leave your torrents seeding for a long time, then it’s in the interest of the swarm to get you the whole file right away so you can be turned into a seeder instead of a leecher who can only share a fraction of the whole file. The top dogs are valuable, so why not get them the file first so they can spread the love? As long as leechers aren’t penalized unnecessarily or harshly I don’t see the problem with giving a few benefits to altruistic seeders.
@Vemrion
That’s basically a description of super seeding mode (as seen in Azureus) right there. The original uploader keeps a record of which peers got which pieces, and once those pieces start to propagate among other leeches the ones with the best apparent share ratios get higher priority on the next round of fresh bits.
It’s the Amway of bandwidth.
This could be a valuable model to follow for pay sites as an alternative to hosting on servers.
Don’t charge by type of media, charge by ratio of download to upload.
Obscure but quality media is more expensive, unless it’s popularity explodes. Then as a sharer, you recoup losses as you upload. The more you upload the more you recoup.
Early new single Britney Spears sharers will rake it in as long as they share, but the bulk of people will pay relatively the same amount, with the stragglers having to pay a little bit more, unless there’s a new surge in popularity after others have pulled Brittney from the share list.
The system feeds the viral wave.
sounds great if ISP will give us as much up as down but till then it might be little tricky to get this going…
what is this “select a version” thing?
is there some sort of documentation about what the deference in the programs is?
Why wouldn’t everyone choose the one on the right side?
http://tv.seas.harvard.edu/
sorry. my bad.
” * Q2: What is the difference between the two download versions?
If you select the left download version, the Tribler client will upload as much as it downloads. This version is “balanced” in the sense that for every piece that you want to download you also have to upload a piece of the same size. If you select the right download version, the Tribler client will optimize the file sharing algorithm to speed up your downloads, minimizing your upload to others. While this improves the speed of your personal video downloads, other users will not be able to benefit from your videos as much as with the left version which then consequently reduces their download speed.”
http://tv.seas.harvard.edu/faq.php
@BRAM dead right
if i was anti-p2p i would set up a private site, invent the ratio, log my members and sell them to the **AA to collect dividends on legal threats
filesharing is NOT about currency of any kind. it is about participation and the small world connection. a bittorrent site is a node which brings people together in a uniquely economic sharing environment
bandwidth is such an unequal thing, only a capitalist could imagine to cash it in
ratio is what keeps bittorrent from proliferating. somebody said above there are a lot of stupid admins, yes there are. there are also a lot of anti-p2p admins
h33t dot com where filesharing is an education
@bram: couple of points.
Some people “share because they care”, but nowhere near the number who share because of ratio requirements.
Most often, a month old file will come down a hell of a lot faster from a private site than a public one.
Taking a release from just 3 days ago, a private site has 409 seeders with 29 leechers. A well known public site has 191 seeders and 203 leechers.
The private site has had 1969 downloads and the public 217.
The private site gets the file to more people, faster.
There’s also files still around on the private site from june 06 with 30-60 seeders.
A popular trilogy from july 1st 06 on a private site has 45 seeds, the same from the public site on december 23rd 06 has 23 seeds.
The private site can continue to get files out with more seeds over a year later.
Of course, we could argue distribution of linux ISOs, but then linux isn’t distributed on any private site I know of.
As for whether ratio is meant to be tracked or not, uploaded and downloaded are part of bittorrent.
Either they are there for a reason, or your protocol has completely useless parts (providing of course you really are bram).
The only reason each peer needs to report how much is uploaded/downloaded is so the tracker has a record of it. It’s certainly not information that has any bearing on the sharing between peers, which rely on their own tit for tat methods.
Either BT is flawed, or ratio tracking is built in and should not be complained about.
One final point. I’ve run a private site for more than 2 years now, and I’ve never once sold upload credit.
But, sales of upload credit are not sales of the users bandwidth to themselves.
Upload credit is used as a currency on private sites, and rarely ever is the total upload credit in the site defined solely by the amount that was actually uploaded.
Extra credit has to be injected into the system to allow it to function properly, as an exactly equal upload/download amount across the site will lead to a stalemate where nobody downloads.
Giving credit to those who donate is only one way of breaking the stalemate.
Many sites have contests running with a credit reward, or will award users who help in other ways with extra credit. Some give credit to all users on a regular basis.
Upload credit does not purely represent a users upload, nor does it mean a user has either uploaded bytes or donated.
Perhaps your criticism should be directed at those who use donations for their own personal gain, rather than any site that rewards those helping to keep the site alive.
I’m pretty sure you had to give your investors some kind of incentive :P
[quote comment="155676"]@Bram : No I guess this is the way to go to reduce excessive leeching on bit torrent.[/quote]
Sounds like a good idea to me
the ratio as it is used on private sites is not built into the protocol. the private site ratio is an aftermarket invention designed to reduce the cost of externalities created by human real world interaction with the protocol
the ratio can be demonstrated to suppress the proliferation of the protocol, reduce potential and nominal swarm sizes, and procreate injustice in the system because not all bandwidth users are created equal
there are better ways to hedge externalities e.g. education, community participation, and system rewarding quality targets (full descriptions, upload frequency, seed to completion rate)
the ratio is a cheap and dirty way to encourage filesharing, and in the end it does more damage than good by limiting the filesharing that happens. the bandwidth currency idea is grown from the ratio. and what inevitably follows a currency? taxs
for now, i leave it for the next man to describe the effect taxation has on an open market. Harvard’s approach is as you would expect from a capitalist business school, is this project perchance somebody’s phd thesis?
regardless, it is great positive step for the protocol when an institution as august as Harvard comes onboard, bodes well for our future
h33t dot com where filesharing is an education
Isnt that just the same as emules ranking system?
untill isp’s give a equal upload to download ratio or at least a close one then i cant see it being too popular, i think it would be a better idea if the bandwidth currency protocol only kicked in on low seeded torrents
[quote comment="157102"]the ratio can be demonstrated to suppress the proliferation of the protocol, reduce potential and nominal swarm sizes, and procreate injustice in the system because not all bandwidth users are created equal[/quote]
I see this comment over and over here, and you know what? Not everyone IS equal, but that’s not “injustice”, it’s reality.
By rewarding the people who upload more, however those people manage to do it, even if it’s unfair (LIFE isn’t fair), ratio-based trackers make the whole system work better for EVERYONE.
People are unequal, yes, but that’s not because a capitalist came by and waved a wand and made it so. THAT’S THE WAY IT IS. GET USED TO IT.
If you’re willing to face reality, you’ll realize that this isn’t the political argument you think it is. I suggest you read a book on economics sometime. Fascinating stuff.
Yah, I know what this means, it means we are moving back to the days of FTP credits. A lot more FAKEZ!!!! You know, people want GTA4. A dude hosts an empty zip file, with 0′z and no compression. People end up getting a lightweight 4 gb file which they downloaded in oh about 10 mins, and the file was padded with zeros so it goes at like 4mbps or so, even on ur 768k cable modem. Meanwhile, uploader gets the credits the downloader spent on him to get it. See how that works? Digital dater theives….
Centralized credit server. This way we could monitor, spot, prevent and/or respond to fakes.
peace
tinkled sunbonnet unmasked spanned pay:supposedly,solubility?
meticulously:obsolescence spun ethically portfolios volunteer sing …
insistence repartition beseech punctuation tense hunched sustained?supermini
mingling:intervention artificially unambitious Anabaptists expediency stifle unsuccessful
I could see currency being used to encourage or discourage seeding. For example, if you have 409 seeders with 29 leechers - not all seeders need to be present for optimal download speeds. To discourage seeding, you lower the credit per megabyte in those instances. At the same time, if you have many leechers with few seeds, you boost the credit per megabyte. A boost in credit could also be applied to real-time data such as streams.
withdraw Blanton.disproportionate Mackinac grabbing framing bangle
Petersburg Lehigh chatters?polytheist trial?…
11 references to this post
Responses are closed
All remaining responses will continue to be archived. Thanks to all who made serious comments.