Choking based on network closeness?
Bill McGonigle <bill <at> bfccomputing.com>
2005-07-11 17:45:16 GMT
I was curious if anyone has experimented with a choking algorithm based
primarily on network closeness, rather than transfer speeds?
For many situations the close peers are going to be the fastest anyhow.
But in some circumstances one might prefer a slower, closer peer than
a faster, farther peer. Those circumstances probably amount to someone
paying for Internet transit at a peering point.
Any deployable implementation would need a heuristic, and maybe user
tunable knobs, as to when to fall back to a fast, far peer.
For certain very popular torrents on a very large 'internal' network,
one can imagine a situation where the percent of the total filesize
transiting an upstream network connection point for download approaches
only 100%, perhaps a very significant reduction from the fastest-peer
I can see a good implementation causing an ISP to prefer BitTorrent
traffic over HTTP traffic, since they're not going to be paying any
transit costs on the local traffic.
Anyway, this doesn't seem particularly clever, so I'm sure somebody's
thought of it before, but I didn't find mention of it in the list
archives. I'd appreciate any stories or pointers.
Bill McGonigle, Owner Work: 603.448.4440