New Torrent Files

[John Reiser]

> Actually for best use of torrent join an existing one rather than setting
> up your own. Big torrents surely work much better than several little
> ones.

It depends.  Transmission delays on trans-oceanic torrents reduce throughput
and change the operating envelope of the protocol, so one torrent per low-delay,
low-contention subnet (such as a continent) often helps.  Undersea fiber is
still expensive, and use by a torrent can skew pricing decisions for next year,
not always in the direction favored by users of torrents.

A torrent with more than several hundred nodes is also slower.  The tracker
can get overloaded, and/or random noise on just a few paths can get amplified
into torrent-wide instabilities and/or inefficiencies.  I observed this
on the duke.edu torrent for FC2T3.  Probably you can see it tomorrow, too.
Even on a fast CPU with plenty of RAM and disk bandwidth, the number of
connections is restricted to something like 100, and on a timescale of
a minute or two, you will only talk to a couple dozen other nodes.
So in practice with moderately reliable links, then there is a bounded
optimum size for a torrent; and today it is likely to be a few hundred
nodes containing a couple dozen seeds.

-- 
John Reiser, jreiser at BitWagon.com