Re: [Sks-devel] Optimum number of peers

[Jeff Johnson]

On Apr 21, 2011, at 12:52 AM, Jeff Johnson wrote:

> 
> Almost certainly someone has modeled gossip protocols, lemme dig a bit.

Caveat:
        This is just the 3-5 minute quick skim search. Take all my comments
        as guesses.

This link
        http://www.distributed-systems.net/maarten/data/papers/2009.icdcn.pdf
has a closed solution (section 5, and figure 7) for a "wireless" gossip model 
that seems
reasonably close to a SKS net.

Note that the results were looking at "optimal" exchange buffer size, not max. 
items on a node.
The model assumed full connectivity, presumably that means all other nodes are 
just 1 hop away.

This link seems pertinent to computing the "optimal" neighbor count:
        http://www.eecg.toronto.edu/~bli/papers/lzhong-globecom10.pdf

Figure 2 would seem to indicate that 5 neighbors is about where the
"fetching rate" maxes out at 100%.

Again this is just random searching and skimming and guessing (while waiting for
downloads to finish *yawn*). YMMV.

hth

73 de Jeff

smime.p7s
Description: S/MIME cryptographic signature