This forum has been archived. All content is frozen. Please use KDE Discuss instead.

Snubbed

Tags: None
(comma "," separated)
jabbervorx
Registered Member
Posts
12
Karma
0

Snubbed

Mon Feb 27, 2006 10:25 am
Hi George,

I experienced a very low down speed for one particular torrent, and looking at peer list I noticed too many "snubbed". Yesterday I downloaded latest KTorrent from SVN, but the problem persisted there also. I grep'ed sources and did following, and now that torrent is much faster.
===
casus@jabbervorx:~/src/ktorrent/libktorrent/torrent$ diff -u peer.cpp~ peer.cpp
--- peer.cpp~ 2006-02-27 01:00:17.000000000 +0100
+++ peer.cpp 2006-02-27 10:48:35.325371944 +0100
@@ -340,7 +340,7 @@

bool Peer::isSnubbed() const
{
- return snub_timer.getElapsedSinceUpdate() >= 60000;
+ return snub_timer.getElapsedSinceUpdate() >= 300000;
}

bool Peer::isSeeder() const
===

Just for your info.


Anton Petrusevich
George
Moderator
Posts
5421
Karma
1

Mon Feb 27, 2006 5:16 pm
Interesting, I will have to experiment a bit with this value.

Are you sure this is not just one big seeder which came along at the right time ?
jabbervorx
Registered Member
Posts
12
Karma
0

Mon Feb 27, 2006 7:32 pm
It's possible. But I definitely see something is wrong with downloading algorithm. Usually looking at peers I see they have 4-6 more percents completed then I have. This is very strange, I did not have such behaviour with qtorrent. My idea was that because of timeouts some peers are getting "snubbed" attribute while they could still be useful.

As a thought: qtorrent worked just perfect downloading files, the problem with it that it is not feature-rich.

If you could document how this thing works in your program and what standard document describes how it should work, I could spend my time thinking with you how to improve things :)


Anton Petrusevich
George
Moderator
Posts
5421
Karma
1

Mon Feb 27, 2006 8:23 pm
jabbervorx wrote:It's possible. But I definitely see something is wrong with downloading algorithm. Usually looking at peers I see they have 4-6 more percents completed then I have. This is very strange, I did not have such behaviour with qtorrent. My idea was that because of timeouts some peers are getting "snubbed" attribute while they could still be useful.


Snubbed means that the peer hasn't sent you a piece in X amount of time (where X is the value you changed). This has it's influence on choking and uploading

As a thought: qtorrent worked just perfect downloading files, the problem with it that it is not feature-rich.

If you could document how this thing works in your program and what standard document describes how it should work, I could spend my time thinking with you how to improve things :)


The code is based on http://wiki.theory.org/BitTorrentSpecification, choking algorithm is based upon some paper I found somewhere. Improvements are allways welcome.


Bookmarks



Who is online

Registered users: abc72656, Bing [Bot], daret, Google [Bot], lockheed, Sogou [Bot], Yahoo [Bot]