Re: Data packages. (with BitTorrent)
Le Sun, Feb 08, 2009 at 07:22:10PM -0600, Lukasz Szybalski a écrit :
> Is there a capability of creating a mirror that would use torrent technology?
> If you have these .deb packages that are big (could you list few, with
> their sizes), I was under the assumption that having torrent mirror
> was possible but rejected because most packages are small <1m. If
> torrent mirror is possible then one could create big.debian.org which
> would be a mirror of big packages that are distributed via torrent
A genome indexed for the BLAST search system takes between 500 and 1000 Mo.
Other databases can be bigger (GenBank – very comprehensive – is almost 100 Go,
I think), or smaller (for instance databases of protein-DNA interactions like
JASPAR or REBASE).
For the general-purpose databases that are updated on a regular basis, torrent
distribution would indeed make sense as we would expect sharp peaks of
parrallel downloads followed by calm periods. For other big datasets – for
instance the human genome – there is less than one update per year, so we would
lose the benefit of a peer-to-peer distribution strategy.
But definitely, torrent is something to keep in mind, especially if
infrastructure makes it transparent. Can debtorrent work with the torrenting
facilities of the Amazon Simple Storage system ?
Lastly, BitTorrent is strictly forbidden in some workplaces (like where I
work), so it will have to be optional.
Have a nice day,
Debian Med packaging team,
Tsurumi, Kanagawa, Japan