[BitTorrent] Dynamic trackers idea

Antonio Augusto khaoticmind at gmail.com
Fri Dec 31 10:44:34 EST 2004



Heya guys,

This is my first message to the group, and also my first attempt to
the P2P development world at all. So it might not be as good as I
first though ;)

Any way...
My idea would be to allow trackers to be created dynamically in the
network.
In my idea the .torrent file would be changed to contain up to five
trackers (one main tracker and 4 "backups") and additional space to a
list of known peers (lets say 50 for now). So, once you start the
download it would normally connect to the tracker, and start the
download. 
If the first tracker did not work it would try to connect to one of
the backups ones. If it did not found any active tracker it'd start
querying the other peers in the list to which is the actual tracker
for this file (and update the backup list as appropriated).

Now, why would the tracker go down?
Well... for any number of reasons. But the thing is that , once a
tracker goes down the peers would communicate among themselves and try
to elect a new tracker. (or maybe more than one to balance the traffic).
I'm pretty sure there are some efficient "electing algorithms" out
there. But i didn't look for any right now.

Any ways, when the new tracker is elected the clients would update
they "backup list" to reflect this change on the topology, so new
clients (that downloaded the new .torrent) would be able to connect
even if the original tracker is down, and latter "resumes" would also
use the backup trackers if the original one is down.


One variation on the topic would be to allow the peers to dynamically
change the topology (i.e. find new trackers) randomly, and not only
when a tracker goes down. This could be useful for example, when a bad
tracker has been chosen (one with low bandwidth), and after that
another peer joined the network, and has better condition to handle
the traffic.
This could also be useful when the network gets to big, so you would
need to redistribute the trackers/create new trackers.

I think that the impact wouldn't be so big for the trackers for a
number of reasons:
1) the tracker elect would host only one file (not thousands of files
like the trackers we have in the current model)
2) the electing algorithm would help to establish that the current
tracker is the one that can best handle the traffic (maybe based on
earlier traffic reports)
3) the elect tracker could have advantages on the queue list of the
peers, so to enable him to achieve better downloads
4) the tracker would change from time to time, so a host wouldn't need
to keep with the traffic all the time

this last assumptions are supposing that the trackers/backups would
exchange a list of any needed info periodically.


Well... i think thats it :)
I also thought about the possibility of making it possible to look for
a file through the network (no need to have separate websites fr
that), but thats another topic i think :)

So... start the critics (constructive ones please ;)





 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/BitTorrent/

<*> To unsubscribe from this group, send an email to:
    BitTorrent-unsubscribe at yahoogroups.com

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/
 





More information about the BitTorrent mailing list