After 90% done, completely slows down.

[Infamous]

I've been having this problem for quite a while now.

Everytime it's done 90%(or close to it) of downloading the file, it will completely come to a slow download speed.

It could be doing 150/kbps to 0.1/kbps maybe 4/kpbs.

Anyway of fixing? Or why it's doing it?

Suggestions / Help is much appreciate - A thank you in advance.

[geezer]

Does this happen even when there are lots of seeds? Or does it only happen when there aren't many seeds? You might be slowing down when you reach the point where only the few seeds have the remaining bits you need. That's just my wild guess.

I'm not an expert (by a long shot) on how super seeding works but I think it causes "young" torrents to be slow when there are peers with extremely low upload caps. I see this phenomenon every week when I download a video blog immediately after it is released. There will be two or three peers which have the highest percentage of the file (by about 5% or so) and they upload next to nothing. They get ahead of the rest of the peers by taking advantage of their opportunistic unchokes. But by uploading almost nothing, they are holding up the release of more data from the seeds which I presume are in super-seeding mode (again, I'm no expert so someone please correct me if I'm wrong). So the whole torrent just crawls until the peers with reasonable upload caps finally get 100% of the file. Then the torrent flies. What are the peers with the super-low upload caps trying to do? Race to see who can get 100% of the file first in spite of the fact that they are holding up the whole torrent? Or are they just idiots who don't know what they are doing with their client settings?

To give a bit more detail, what I noticed about these peers is that they unchoke my client, my client queues some requests, they never actually send any data, and then they choke me again. This happens over and over. Very strange. Of course, my client is perfectly happy to send data to them (opportunistic unchoke). Which is why they've got at least 5% more of the file than the rest of us "cooperative" peers.

Sorry for ranting! Sorry also if I'm misrepresenting how this stuff works.

[DutchDude]

I'm not an expert (by a long shot) on how super seeding works but I think it causes "young" torrents to be slow when there are peers with extremely low upload caps

It has nothing to do with super-seeding as i understand it... Superseeding means that a torrent is initiated (i call him the initiator). That initiator can use super-seed mode. This mode is used when a person has low upload bandwidth. It's goal is to ensure that any client is supplied with as few packets as possible to complete the torrent. The missing packets are then requested by means of contacting other leechers.... The swarm (not including the initiator) has all required parts among them to complete the torrent, without having to connect to the original seeder/initiator. This method ensures that the seeder/initiator in super-seeding mode is not robbed of his entire upload bandwidth for a single torrent.... Besides this it is possible that some parts were given to peers with low upload capacity....requiring you to wait for a (long) while to get them.

What's happening here is that probably a few parts are either no longer available, or the peers that have them are busy enough for him to have to wait for the remaining parts.... Remember... any given peer can be a seeder of any given number of torrents..... and the upload bandwidth is divided unevenly among all running uploads, so he just have to wait for the torrent to complete...

BTW... in the bottom of the screen there is a 'Pieces' and 'Files' tab... which could help you to see if a specific file is available and what parts are rare... In the General tab you can see if the torrent as a whole is completely available (If the number is <1 then the torrent is not completely available at this time)...

[geezer]

Thanks for explaining. I shouldn't have assumed that the behavior of those three stingy peers would have any effect on the super seeding algorithm the seeds were using. If I understand correctly, the torrent would be slow whether those three peers were stingy or not, just because the seeds are uploading data very slowly. Their stinginess just gets them a few more % of the file than the rest of us have at any given time. I don't have any problem with that as long as their behavior doesn't have any effect on the rate at which the seeds release new data.

[DutchDude]

Thanks for explaining. I shouldn't have assumed that the behavior of those three stingy peers would have any effect on the super seeding algorithm the seeds were using. If I understand correctly, the torrent would be slow whether those three peers were stingy or not, just because the seeds are uploading data very slowly. Their stinginess just gets them a few more % of the file than the rest of us have at any given time. I don't have any problem with that as long as their behavior doesn't have any effect on the rate at which the seeds release new data.

I get the distinct feeling that you mix-up seeds and leechers (together called peers... which are referred to as a swarm).

The super-seeder will give out any part that is not present within the swarm to any client, because at this point there may not be any seeds (some peer that has 100% of the file). Now it is possible that there are 3 peers who among them selves have all parts, but none single peer has a complete torrent. This means that the super-seeder does not give out any more parts to the swarm. If those peers have low bandwidth (and in fact are uploading to more then 1 torrent) it is very well possible to get low downloads from a torrent.... The more seeds there are (and the less they are busy downloading themselves or uploading to other clients) the higher your download will be...

(Just keep in mind.... the seeds may not use BT only.... they could also be running emule or some other P2P program which also would consumate some part of their upload bandwidth)

Dispite this all... it all comes down to a single question... Are you patient enough ?

[geezer]

When I said "peer" I meant a client which doesn't yet have 100% of the data. When I said "seed" I meant a client with 100% of the data. What I definitely did wrong was ascribe WAY, WAY too much magic to the super-seeding algorithm because I've only read very brief, sketchy descriptions of how super-seeding works. I stand gratefully corrected on that point. I should not have posted on this topic until I'd found a better description. I apologize to you and to the original poster.