Wow, you're right, that's a lot of seeds. I didn't know there were any torrents that well seeded. So by that measure, assuming BTJunkie is comprehensive, there are precisely 37 torrents in the world with "a lot" of seeds. And all the millions of other torrents don't.
But regardless, total number of seeds probably isn't the most relevant measure. Rather, the key indicator of download speed is seeder/leacher ratio: the more uploaders there are relative to the number of downloaders, the faster the average download will go. A 1:1 seeder/leacher ratio means 2 uploaders for every downloader (with one the uploaders also actively downloading). A 2:1 ratio means 3 uploaders/downloader, etc. The higher the ratio, the more upload capacity each downloader can draw upon. Here are the seeder/leacher ratios of the top 5 most active torrents in the world: Title Seeders Leachers Ratio Belenix OpenSolaris Install DVD 195815 87029 2.24 Ubuntu 7.10 Desktop AMD64 ISO 60472 31018 1.94 No Country For Old Men 2007 DvDRip 21049 14949 1.41 The Mist[2007]DvDrip[Eng] 20363 62894 0.32 Hitman[2007][Unrated Edition] 18294 20895 0.88 Not bad -- the top download has more than 3 uploaders for every download. Cool! But it quickly goes down from there. (A better analysis would sort BTjunkie by seeder/leacher ratio, perhaps filtering out any torrent with less than 1000 seeds, but this'll do in a bind.) So the most active torrents in the world are eeking by with a few more uploaders than downloaders. But let's remember that the typical connection is asymmetric: you can typically download something like 4x faster than you can upload. This means that to saturate your downlink, all things being equal, you need four other uploaders to saturate their uplink. So a ratio of less than 4:1 means you can't, on average, download as fast as you'd like. In other words, any torrent with a seeder/leacher ratio of less than 3.0 will fail to fill the average downloader's pipe, and thus will download slower than a well-provisioned webserver. So you might think "well duh, a massive server will always download faster than P2P, but it'll cost through the nose!" But this isn't true -- if you had more than 4x as many uploaders as downloaders, then you'd start to download as fast as a webserver, except for "free" from peers. Unfortunately, this situation is pretty rare with Torrents. Not that it doesn't happen -- especially on commercial content with well provisioned seeds (that are being paid for, and aren't afraid of the Man coming down hard on them). But it is uncommon, if not rare. This is why I argue that bittorrent isn't about performance, it's about piracy. And that's fine -- there are a lot of pirates out there, and performance isn't their top concern. Building a tool for them makes a lot of sense. But there are other P2P networks out there that *are* focused on performance. I'm guessing Bittorrent DNA is one. I know Red Swoosh is another. And to give an example of what we mean by performance, at Swoosh we view any download with less than a 20:1 seeder/leacher ratio as a temporary anomaly. Meaning, 20:1 is our *low* end -- we're designed to operate with ratios in the hundreds or thousands or more. There are real, fundamental differences between protocols that are designed for seeder/leacher ratios of 2.0 and less (Bittorrent), and 20.0 or more (Red Swoosh). And that's the difference between optimizing for pirates or performance. Both are interesting problems. But they're very different problems, with very different solutions. -david Bill Mccormick wrote: > Empirically "alot" of seeds would be > 10000. > > Some torrents with a lot of seeds today are: > > OpenSolaris install DVD 195000 > Ubuntu 7.10 Desktop 60000 > No Country for Old Men 21000 > The Mist 20000 > > etc. > > You can see some rough statistics at torrent search sites like btjunkie.org. > > BT clients won't always seed if there is a high enough share ratio. > (Excuse the odd terminology, this is from the utorrent UI.) So > that's one reason not everyone seeds. > > Another one is just the clutter in the BT client. i.e. Every active > torrent takes up a line in the UI. > > And a third reason is that every seed takes up some of your uplink bandwidth. > > I tend to think that most users are not particularly concerned with > the piracy aspect, even after the recent court decision in the states > in favour of the RIAA. Many countries don't provide the same level > of copyright protection to content creators as the U.S. does. > > But I will try and get some data one of these days.... 8-) > > Bill. > > > > On 3/11/08, David Barrett <[EMAIL PROTECTED]> wrote: >> What's "a lot" of seeds? If a million people download a particular >> file, is a lot a hundred? A thousand? The question is, why not a >> million seeds? (The answer to that question is "because then a million >> people are vulnerable; better to only have a few high-risk people stick >> out their necks.") >> >> My point is only a tiny fraction of total downloaders actually stick >> around as persistent seeds. Maybe 1%? 0.1%? That means somewhere >> between 99% and 99.9% of potential swarm capacity is thrown away in >> order to preserve the pirates' anonymity. >> >> That's cool -- I'm not dissing bittorrent. It's truly brilliant. It's >> just designed for piracy. That's fine -- the best technology often is. >> But let's not pretend otherwise. >> >> -david >> >> PS: The one possible explanation for the Bittorrent design that doesn't >> implicate it in catering to pirates is it's the only design that works >> in an open-source, multi-implementation environment where you generally >> can't trust your peers: tit-for-tat is built for a paranoid world and >> thus can survive when that world is reality. I'm sure this design goal >> plays a factor. But I suspect its legal protections are a much bigger >> reason for its widespread adoption. >> >> Bill Mccormick wrote: >>> On 3/11/08, David Barrett <[EMAIL PROTECTED]> wrote: >>>> I agree a well-seeded torrent can be pretty quick. But the >>>> protocol/clients/users only seed for a very limited time, or not at all. >>>> The result is most torrents are poorly seeded, and thus slower than >>>> downloading from a well-provisioned webserver. Said another way, >>>> Bittorrent generally sacrifices speed in order to protect pirates. >>>> >>> >>> --> I'll check that one the next time I'm downloading a smaller >>> torrent. I have the impression that alot of seeds are long term >>> persistant. But no data to back it up yet. >>> >>> Bill >>> _______________________________________________ >>> p2p-hackers mailing list >>> [email protected] >>> http://lists.zooko.com/mailman/listinfo/p2p-hackers >>> >>> >> _______________________________________________ >> p2p-hackers mailing list >> [email protected] >> http://lists.zooko.com/mailman/listinfo/p2p-hackers >> > > _______________________________________________ p2p-hackers mailing list [email protected] http://lists.zooko.com/mailman/listinfo/p2p-hackers
