AW: mulitcast copy or snowball copy
mitchskin at comcast.net
Mon Aug 18 13:40:59 EDT 2003
On Mon, 2003-08-18 at 09:50, Donald Becker wrote:
> This is costly. "Open loop" multicast protocols work by having the
> receiver track the missing blocks, and requesting (or interpolating)
> them later. Here you are discarding that information and doing much
> extra work on both the sending and receiving side by later locating the
> missing blocks.
Some possible google terms include: reliable multicast, forward error
There's an ietf working group on reliable multicast that wasn't making a
whole lot of progress the last time I checked. At that time, I recall
there being some acknowledgment-based implementations as well as one
forward error correction-based implementation using reed-solomon codes,
from an academic in Italy whose name I forgot.
It's been a little while, but when I looked at the code for that
FEC-based reliable multicast program (rmdp?) I think it could only
handle pretty small files. My understanding is that FEC-based
approaches should scale better in terms of the number of receiving
nodes, but the algorithms can be very time/space intensive. There's a
patented algorithm from Digital Fountain that's supposed to be pretty
efficient (google tornado codes, michael luby, digital fountain) but I'm
not aware that they have a cluster-oriented product. My impression of
them was that they were pretty WAN-oriented.
If I was less lazy I'd give some links instead of google terms, but
hopefully that's some food for thought.
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
More information about the Beowulf