jMule

[bill]

I am starting to look at different methods for matching Peers in a P2P system and am having a difficult time finding documentation on how existing protocols (jMule,Kazza,Gnutella) determine which is the best server to select for a client requesting a download of a file that may exist on more than one server.

Does anyone here know how jMule performs this selection? Does it use a simple metric of ping time like Napster or something more sophisticated like "amount of data currently left to serve" or "number of concurrent downloads w.r.t bandwidth available" or something like that...

Thanks in advance for any info you might be able to provide

[emarant]

jMule is cable of not only use one protocol to transfer files, but currently
jMule uses the edonkey2000 multisource file transfer protocol, like eMule, edonkey2000, mldonkey, overnet and other edonkey2000 compartible clients, too.
Those *clients* act all as file server in the network.
Request of files:
simply ask all peers haveing parts of the file needed, for a transfer. The uploading peer(server) keeps all requesting peers in a queue until it has a free upload slot. If a slot becomes free the server takes a peer from queue and than the downloading peer trys to get the rarest parts of the file available on all peers, the server can provide.
So jMule work is: find all source for a file asking eMule compartible clients down-/uploading the file, edonkey2000 server and maybe in future overnet or overnet like network peers.
Than request file form sources have needed parts.
If a peer gives jMule an uploadslot, only thing jMule selects are the parts of a file to download from this peer.
The decision which client can download form a server is always made by the server.
Any uploading peer can stop uploading using his own policy, like give everyone only x bytes per queue cycle or reward some kown peers for their previous upload and give them more than unkown/freeloading peers.