BitTyrant

Has integrating BitTyrant like behavior into µTorrent been considered? In my case, when there are relatively few seeders, BitTyrant downloads torrents at least twice as fast as µTorrent does. In general µTorrent is a far superior application. But the increased download speed of BitTyrant is so dramatic that, for me, it trumps all the benefits of using µTorrent.

Has BitTyrant like behavior been ruled out or would the developers consider adding it if there were conclusive evidence that it's overall effect was beneficial?

Is there real world evidence that BitTyrant use actually harms swarms? I have read that it could hypothetically harm swarms when a user is downloading many torrents at once. But I have also read that it more efficiently distributes bandwidth, resulting in greater download speeds for everyone.

- Ian

P.S. For your reference, my upload speed is usually capped at around 220kB/s by the Autospeed plugin.

It's obvious that hit-and-runners using ultra-aggressive download-side BitTyrant behavior WILL wreck torrent swarms if they leave very shortly after reaching 100% complete.

uTorrent already has allow additional upload slots if uploading slowly.

It also seeks to tit-for-tat with other peers.

It even REDUCES used upload slots below max if you're trying to use too many at once for your upload speed.

Optimizing seeding/uploading behavior is more important than download behavior.

I don't think it's obvious that widespread BitTyrant use harms swarms. Is there any real world evidence that BitTyrant does harm to swarms?

Anecdotally, when I download with BitTyrant from swarms with a low ratio of seeders to leechers my downloads usually complete with a share ratio around 1.2:1. That's not conclusive evidence that widespread BitTyrant improves swarm health but I think it's worth investigating.

The developers of BitTyrant think it generally improves swarm health and the people at the CS department of the UW are pretty smart. The Wikipedia article on BitTyrant has a relatively simple explanation of why they may be right.

> widespread BitTyrant use harms swarms

I think Switeck was suggesting it depends on how you use BitTyrant. If you leave after a torrent is completed and not contribute upload bandwidth, then the swarm is negatively affected. If you stay around and seed indefinitely, I doesn't matter much. Just from personal experience, a lot more people run after a torrent is finished than not. Ever since µTorrent started showing total finished and number of seeds left on some torrents, its really depressing to see how selfish most of the torrent users are.

People leaving the swarm earlier doesn't necessarily result in lower upload ratios.

Imagine a swarm with one seeder and two leechers. Say the seeder has an upload capacity of 200 kB/s, one leecher has an upload capacity of 200 kB/s and one leecher has an upload capacity of 50 kB/s.

If the seeder uploads to both leechers at 100 kB/s, the faster leecher will upload at 100 kB/s to the slower leecher and the slower leecher will upload at 50kB/s to the faster leecher.

In this scenario the faster leecher gets a download speed of 150kB/s and the slower leecher gets a download speed of 200 kB/s.

But if the seeder uploads to the faster leecher at 200 kB/s, the faster leecher can upload to the slower leecher at 200 kB/s, also. In this secario both leechers get a download speed of 200 kB/s. The fast leecher does leave the swarm earlier, but not at the expense of the slow leecher.

These scenarios are artificial but I think they demonstrate that, in comparison to BitTyrant, µTorrent may disadvantage fast uploaders without benefiting slow uploaders.

Also, upload-to-download ratios aren't NEARLY as important as availability of a torrent after you leave. Many torrents "die" because the remaining seed that's already uploaded 4+:1 ratio leaves thinking they've done their part.

With people setting their clients to download sequentially, the problem is even worse.

I understand that availability is important, but can you explain how availability would be a problem if upload ratios weren't a problem? As far as I can tell availability problems stem from upload ratio problems. (Unless, of course, the first seeder never actually uploads 100% of his torrent.)

1st seeder faces a rash of sequential downloaders...of the first 10 downloaders, 5 of them insist on downloading from beginning to end. They demand the first piece/s of the torrent from the seed over all others. So the seed has to upload the same pieces over and over again! The other peers that are trying to play fair have nothing to offer to these downloaders...and quickly get the early parts but can get "crowded out" of the seeder's limited upload by the greedy sequential ones. It's not uncommon for a seed to upload more than 4x the torrent size before the first additional seed appears.

Beyond raw ratio is piece boundary issues. Unless a piece is completed and checked, it's worthless to other peers. This can become a nasty problem when piece size is 4 MB. BitTyrant's approach of uploading to potentially 20 people at once instead of limiting itself to only 1-5 like uTorrent typically does means there's far more likely to be 20 partial pieces out there than 2-5 peers with MULTIPLE completed pieces.

OK, so I see how sequential downloading damages the health of a swarm. (Though I don't think that's a problem specific to BitTyrant.)

In the second scenario you bring up I see potential benefits to both BitTyrant and µTorrent's behavior.

With only slow peers BitTyrant would upload slowly to many slow peers and µTorrent would upload quickly to a few slow peers.

BitTyrant would slowly create a greater number of productive uploaders. (Coincidentally, I think this behavior would tend mitigate the damage done by a few sequential downloaders by increasing peer diversity.)

On the other hand, µTorrent would quickly create a lesser number of productive uploaders.

Modeling which strategy works best is pretty complex. So what I'd like to know is is there real world evidence that shows one strategy to be more effective and if there isn't would µTorrents developers be open to including BitTyrant style behavior if there was conclusive evidence that it was more effective?

From the testing BitTyrant's developers did, they seem to have concluded that BitTyrant swarms are significantly more healthy.

I don't know how extensive or thorough their testing was but, for what it's worth, I would be willing to help with testing in any way I can if µTorrents developer's are open to integrating BitTyrant behavior into their program.

Is it better to use 10 megabit of bandwidth to upload 1 MegaBYTE of a file spread between many peers...or 8.5 megabit of bandwidth to upload 1 MegaBYTE of a file to only 1-3 peers?

Is it worth considering the second scenario (considering how most people have rather limited upload) even if it might reduce download speeds?

Or should we also consider the other direction, where even reducing the upload speed limit we use might increase download speeds?

I don't know the answers to your questions, but I think it's worth considering any changes that you believe will improve swarm health overall.

The situation we have now is probably a race to the bottom.

YouTube settings so bad that I can fully understand ISPs banning BitTorrent completely.

More hostile ISPs = worse settings, as people try to use in uTorrent what they think they have.

People who believe the 'fastest' download speeds come from setting global upload speed to 1 KB/sec and then upload at unlimited speed once the download completes.

Anecdotally, when I download with BitTyrant from swarms with a low ratio of seeders to leechers my downloads usually complete with a share ratio around 1.2:1. That's not conclusive evidence that widespread BitTyrant improves swarm health but I think it's worth investigating.

Actually, that can still mean the reverse.

uTorrent would likely have a higher upload-to-download ratio by the time it's done, meaning it's probably helped the swarm more by the time it's done.

BitTyrant idea for strategic peer is not new, just emphasized. All bittorrent clients actively search for this potential peers.

I don't know if we choke peers, would they give us more?

Tit for tat schemes surely slow downs leechers, but also slowing down data distribution. I prefer to give upload freely (as long as don't hurt my download significantly), and if my peer turns out to be a leecher - that's his/her choice.

I want to have a button to disable tit for tat schemes, but optimistic un-choking already serve my purpose, even better - because it unchoke peers that really need the data. BitTyrant do more than that, it unchoke to peers that have high upload (info from scape).

utorrent faster download with ipfilter.dat or setting - http://forum.utorrent.com/viewtopic.php?pid=468434#p468434