[00:36:05] *** goussx_ has joined #bittorrent
[00:53:49] *** goussx has quit IRC
[00:53:49] *** goussx_ is now known as goussx
[01:17:51] *** klapaucjusz has quit IRC
[01:41:53] *** andar2 has quit IRC
[02:15:02] <The_8472> another reason why streaming (or anything sequential) is horrible...
[02:15:10] <The_8472> someone is downloading things mostly in order
[02:15:16] <The_8472> several people are
[02:15:24] <The_8472> and the rest is downloading in reverse order
[02:15:54] <The_8472> which means the first file will finish last for anyone who doesn't use file priorities/streaming/whatever
[02:17:24] <BentMyWookie> streaming for a protocol designed for downloaders to share what they've already downloaded with other downloaders is clearly a horrible idea
[02:17:54] * BentMyWookie can't wait to see what swarms are like once utorrent supports this
[02:18:08] <Switeck> "pure" Seq. DLers that download only the next piece in a row are even worse -- they never tit-for-tat with anyone.
[02:18:29] <The_8472> that's not the case in what i'm observing
[02:18:36] <The_8472> they're just putting higher priorities on the first files
[02:18:37] <Switeck> my bad, never with other Seq. DLers I mean
[02:18:57] <The_8472> and that alone is already enough to make everyone else download in reverse order
[02:19:52] <Switeck> from a seed's point of view, any request for a duplicated (by peers) piece comes at an opportunity cost of uploading a rare piece instead.
[02:19:52] <The_8472> the sequential downloaders still seem to be requesting a few pieces from the files towards the end, but not many
[02:21:41] <Switeck> "The most important decision that we need to bake into the user experience is deciding the point at which a file is ?streamable? " ...In short, convincing the users that many times they can't stream.
[02:21:55] <The_8472> i know swarms that are seeded by several 100Mbit seedboxes but get downloaded via RSS by thousands of people as soon as they're released. and the seedboxes move on to other torrents after a few hours
[02:22:33] <The_8472> so the swarm appears fast initially due to the many seedboxes, but if people stream from them they're basically abusing seed bandwidth
[02:22:41] <The_8472> this will be fun...
[02:22:56] <Switeck> I was told awhile back that uTorrent didn't differentiate between the relative value of seeds
[02:23:10] <Switeck> a "seed" with global upload speed set to 1 KB/sec still adds 1 to advailability
[02:23:17] <Switeck> availability
[02:23:26] <The_8472> well, you can't really measure that remotely
[02:23:44] <The_8472> whether it's a good seed or just a placebo
[02:23:57] <Switeck> perhaps not, but it's crucial in knowing how well the torrent swarm can tolerate streaming
[02:24:32] <K`Tetch> maybe we should test
[02:24:36] <Switeck> Many seeds may have excellent upload speeds but be connected to 50-200 peers but only uploading to 4.
[02:24:41] <The_8472> i guess you can only go by what you see from other peers and the skew in the piece availability distribution
[02:25:21] <K`Tetch> build up a custom swarm to test and measure this, purpose made tracker etc
[02:25:28] <Switeck> what if it goes by how many seeds it "thinks" the swarm has? :P
[02:26:02] <The_8472> K`Tetch, you need at least ~100 nodes to emulate realistic swarm behavior
[02:26:22] <The_8472> that would be a horrible measure
[02:26:43] <The_8472> you can only go by live stats observed among the connected peers
[02:26:44] <K`Tetch> thers 50 of us here, i reckon between most of us we could come up with another 10 more each
[02:26:45] <Switeck> I'd say ~100 is the minimum number of nodes to consider streaming.
[02:28:26] *** Andrius has quit IRC
[02:28:35] <The_8472> it doesn't matter. take my scenario above. fresh swarm, well seeded temporarily by ~10 seed boxes, 1000 peers. conditions seem fine for streaming because the seedboxes are pumping lots of bandwidth into the swarm
[02:28:38] <The_8472> so people stream
[02:28:45] <The_8472> and then BAM ... the seeds disappear
[02:29:16] <The_8472> and suddenly you're sitting there with many peers who have sequentially downloaded (to some extent) the same stuff
[02:29:35] <Switeck> there's also the knock-on effect of some doing it
[02:30:01] <Switeck> other peers not intending to Seq. DL will download from those that do...and they too will get predominately the "early" pieces.
[02:30:56] <Switeck> While this may initially help streaming out, with higher availability of the early pieces...it means that's what the swarm's doing instead of trying to get ALL the pieces evenly.
[02:31:08] <The_8472> not necessarily
[02:31:27] <Switeck> depends somewhat on how little download BW a peer has
[02:31:30] <The_8472> as long as the streaming peers do at least some rarest first the non-streaming ones will basically ONLY download those pieces
[02:31:49] <The_8472> more or less completely ignoring the sequential parts
[02:32:13] <Switeck> What if the seq. parts appear first?
[02:32:23] <The_8472> huh?
[02:32:34] <Switeck> Does a torrent client typically drop a request for a common part to switch to a rare piece from the same peer?
[02:32:43] <K`Tetch> should do
[02:32:51] <The_8472> depends on the client
[02:33:07] <The_8472> there are some conflicting goals... finishing pieces so you can share them and starting rarest ones
[02:34:06] <Switeck> Also, piece size may play a role...at 256 KB or 512 KB, things may get by ok. At 2 MB or 4 MB...problems can compound.
[02:34:46] <The_8472> good point
[02:35:10] <Switeck> much harder to get 4 MB pieces in a fractured swarm with other peers with dissimilar goals.
[02:35:17] <The_8472> 4MB pieces and a low aggregate throughput means the swarm takes aeons to fix any skews in the distribution
[02:35:48] <Switeck> just to get their first piece to share...may mean they resort to downloading an "early" piece that has increasing number of sources
[02:36:20] <Switeck> once they get that, that early piece will be competing with rarer pieces to be uploaded via their no doubt limited upload.
[02:36:39] <The_8472> anyway streaming certainly is not GOOD for swarm health. thus it means that streaming sacrifices performance for everyone for the benefit of the few.
[02:37:06] <Switeck> do you think it sometimes or always slows down completion times for peers?
[02:38:08] <The_8472> depends. i guess it wouldn't really have an impact on completion times on heavily seeded torrents. but it'll still introduce heavy skew into the piece distribution... if it persists then it might lead to issues later in a swarm's lifecycle
[02:38:29] <Switeck> kinda like it "paints itself into a corner"?
[02:38:30] <The_8472> i mean there are situations where streaming is tolerable
[02:38:40] <The_8472> yeah
[02:39:32] <The_8472> the issue is that normal, not overseeded swarms may allow a few users to stream without any immediate impact
[02:39:40] <The_8472> but a) what if many users do it
[02:40:06] <The_8472> b) it's hard to predict if it'll have an impact on the lifetime of a torrent
[02:41:05] <The_8472> so at the very least the conditions on streaming should be conservative, since an automated swarm health assessement can hardly account for all conditions that occur in the wild
[02:41:05] <Switeck> firewalled peers/seeds play a big role in the lifetime of a torrent.
[02:41:32] <Switeck> Even a high number of seeds may be of limited value if many/most are firewalled.
[02:42:09] <Switeck> but it's rare that trackers even sort out firewaled/unfirewalled seeds/peers
[02:43:20] <The_8472> i'm not sure what kind of impact streaming would have on a partially bipartite graph
[02:43:46] <The_8472> well, it's not really necesary these days. PEX works quite well in connecting those who really are connectable
[02:44:07] <The_8472> it's more about the clustering that it introduces
[02:45:39] <Switeck> Such as cluster subdomains on a torrent flagged as private?
[02:49:26] <The_8472> well, those are far far worse
[03:06:50] *** ajaya has quit IRC
[03:08:44] *** dandon has quit IRC
[03:27:45] *** GTHK has quit IRC
[03:28:07] *** GTHK has joined #bittorrent
[03:40:01] *** GTHK has quit IRC
[03:45:40] *** GTHK has joined #bittorrent
[04:08:06] *** The_8472 has quit IRC
[04:08:27] *** The_8472 has joined #bittorrent
[04:23:21] *** init0 has quit IRC
[04:24:43] *** init0 has joined #bittorrent
[04:43:55] *** bt42 has quit IRC
[04:46:31] *** bittwist has joined #BitTorrent
[05:08:03] *** goussx has quit IRC
[05:12:59] *** goussx has joined #bittorrent
[05:45:02] *** rrr_ has quit IRC
[05:51:54] *** andar has quit IRC
[05:56:37] *** Switeck has quit IRC
[06:38:06] *** choykloun has quit IRC
[06:51:41] *** MassaRoddel has quit IRC
[06:55:06] *** GTHK has quit IRC
[06:58:12] *** andar has joined #bittorrent
[07:55:23] *** bt42 has joined #BitTorrent
[08:15:27] *** bittwist has quit IRC
[08:56:40] *** bittwist has joined #BitTorrent
[09:16:30] *** MassaRoddel has joined #bittorrent
[09:16:31] *** bt42 has quit IRC
[09:46:50] *** codas has joined #bittorrent
[09:47:30] <codas> hi! how do seedboxes work? do i need a fast internet connection? or can i make use of a seedbox that is already installed in such a network?
[09:50:15] <chelz> both
[09:50:54] <chelz> just needs to be a reasonably powerful machine with a nice amount of hdd space and a good internet connection
[09:53:35] <codas> yep
[09:53:43] <codas> what if i dont have a good internet connection?
[09:54:51] *** _rafi_ has joined #bittorrent
[10:01:28] <chelz> you can pay to rent a server
[10:07:00] <codas> i see
[10:07:09] <codas> do you have links ?
[10:15:39] <codas> another question: do you know where can i find documentation about the heuristics used by seeds to decide which swarms to prioritize, when many swarms are to be served?
[10:24:41] <jnpplf> As for the seedbox, that will depend where you're based
[10:24:53] <jnpplf> No idea about docs
[10:25:25] <chelz> the src of some oss clients might suffice
[10:25:34] <codas> oss ?
[10:25:51] <jnpplf> Open source
[10:26:14] <codas> thanks!  so, there is no standard for that across clients, right ?
[10:26:23] <codas> i'm curious to check vuze policy
[10:26:52] <codas> the doc would be helpful to understand the rationale behind some policies ....
[10:27:23] <jnpplf> Don't expect rationales to ever be explained ;)
[10:28:05] <codas> :)  for instance, why the seeds select a maximum number of swarms to serve concomitantly ?
[10:28:53] <codas> i guess it is to guarantee a minium bandwidth to each swarm .... but why is that so crucial ?
[11:34:52] *** bt42 has joined #BitTorrent
[11:54:50] *** bittwist has quit IRC
[12:28:14] *** chelz has quit IRC
[12:31:48] *** muetze has joined #bittorrent
[12:32:00] <muetze> hello
[12:33:45] <muetze> as I have never done before I would like to knew how to help seeding some Files I already have received over another channel than Bittorrent
[12:35:11] <muetze> and second which commandline only bittorrent client schould i use at best for help seeding large Files under debian?
[13:07:15] *** PN has joined #bittorrent
[13:07:55] *** ProperNoun has quit IRC
[13:10:47] *** andar has quit IRC
[13:10:47] *** Mazon has quit IRC
[13:10:47] *** kjetilho has quit IRC
[13:10:48] *** muetze has quit IRC
[13:10:48] *** goussx has quit IRC
[13:10:49] *** softwareelves has quit IRC
[13:11:33] *** muetze has joined #bittorrent
[13:11:33] *** andar has joined #bittorrent
[13:11:33] *** goussx has joined #bittorrent
[13:11:33] *** Mazon has joined #bittorrent
[13:11:33] *** kjetilho has joined #bittorrent
[13:11:33] *** softwareelves has joined #bittorrent
[13:14:20] *** bittwist has joined #BitTorrent
[13:15:02] *** stalled has quit IRC
[13:15:48] *** stalled has joined #bittorrent
[13:21:41] *** kwinz2 has quit IRC
[13:21:58] *** bt42 has quit IRC
[13:25:28] *** choykloun has joined #bittorrent
[13:25:40] *** kwinz2 has joined #bittorrent
[13:25:45] <choykloun> ah .. new years.. might mean that i finally have time to impelemnt my new lookup algorithm for dht ..
[13:28:03] <choykloun> combintion of binary search and hash tables to be able to lookup both individual IDs and deltas without any max limit and as close to linear time as possible
[13:29:12] <choykloun> i just dont like doing things the way all the other kids do it :D
[14:04:05] *** kwinz2 has quit IRC
[14:05:22] *** rrr_ has joined #bittorrent
[14:21:21] *** TheSHAD0W` is now known as TheSHAD0W
[14:24:05] *** goussx has quit IRC
[14:43:26] *** muetze has left #bittorrent
[14:56:18] *** n215_ has quit IRC
[15:13:32] *** kwinz2 has joined #bittorrent
[15:38:44] *** Snoopotic has quit IRC
[15:56:45] *** bt42 has joined #BitTorrent
[16:05:07] *** kwinz2 has quit IRC
[16:15:48] *** bittwist has quit IRC
[16:28:21] *** PN is now known as ProperNoun
[16:31:56] *** K`Tetch has quit IRC
[16:39:50] *** K`Tetch has joined #bittorrent
[16:40:30] *** kwinz2 has joined #bittorrent
[16:52:23] *** codas has quit IRC
[17:07:23] *** codas has joined #bittorrent
[17:15:14] *** Switeck has joined #bittorrent
[18:22:51] *** andar2 has joined #bittorrent
[18:44:04] *** codas has quit IRC
[19:21:22] *** Switeck has quit IRC
[19:22:50] *** kwinz2 has quit IRC
[19:22:54] *** _rafi_ has quit IRC
[19:59:59] *** swinokur has quit IRC
[20:02:20] *** swinokur has joined #bittorrent
[21:03:33] *** JudgeSHAD0W has joined #bittorrent
[21:24:17] *** bittwist has joined #BitTorrent
[21:42:02] *** JudgeSHAD0W has quit IRC
[21:42:03] *** bt42 has quit IRC
[21:42:03] *** kjetilho has quit IRC
[21:42:03] *** andar has quit IRC
[21:42:04] *** Mazon has quit IRC
[21:42:04] *** rrr_ has quit IRC
[21:42:04] *** softwareelves has quit IRC
[21:43:25] *** GTHK has joined #bittorrent
[21:53:18] *** kwinz2 has joined #bittorrent
[21:53:18] *** rrr_ has joined #bittorrent
[21:53:18] *** softwareelves has joined #bittorrent
[21:55:16] *** JudgeSHAD0W has joined #bittorrent
[21:55:16] *** andar has joined #bittorrent
[21:55:16] *** Mazon has joined #bittorrent
[21:55:16] *** kjetilho has joined #bittorrent
[21:55:41] *** stalled has quit IRC
[21:56:24] *** stalled_ has joined #bittorrent
[21:59:41] *** stalled_ is now known as stalled
[22:00:49] *** stalled has quit IRC
[22:01:29] *** stalled has joined #bittorrent
[22:15:06] *** rrr_ has quit IRC
[22:17:34] *** kwinz2 has quit IRC
[22:20:27] *** rrr has joined #bittorrent
[22:26:47] *** kwinz2 has joined #bittorrent
[22:44:05] *** burris has quit IRC
[22:44:27] *** burris has joined #bittorrent
[23:34:08] *** MassaRoddel has quit IRC





top