On Wed, Dec 13, 2006 at 08:35:56PM +0000, Michael Rogers wrote: > toad wrote: > > Most of the time they are trying to send swaps out, which get rejected > > because the node they reach is already locked, because it itself is > > sending a swap out. Executing swaps doesn't take long; it's the 6 hops > > random routing to find a partner who isn't already locked, the reply, > > the confirmation and the completion that take time; but maybe we can > > significantly speed this up by queueing accepts, so that nodes don't > > fail to swap most times? > > Sorry, I still don't understand how it's possible to speed things up > significantly if (a) each node can only participate in one swap at a > time, and (b) nodes are already spending most of their time > participating in swaps (which is indicated by the fact that most nodes > are locked). If swap requests are arriving faster than the node can > serve them, won't the queue just get longer and longer?
The problem is that a lot of the time the node is actually not engaged in a swap, but waiting for a search to find somebody to swap with who isn't locked. 5/6ths or so of all swap attempts result in failure due to the other side already being locked. If a node sends a request out, it is locked until that request completes, is rejected, or times out. > > Cheers, > Michael
signature.asc
Description: Digital signature
_______________________________________________ Devl mailing list Devl@freenetproject.org http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
