If someone does model this I think it would be interesting to try "random" 
caching. What happens if say 10% of traffic passing through a node is 
randomly selected for caching.  I suspect that the level of redundacy is 
*too* high in freenet causing stuff to be pushed out of the store quickly.

Matt
--

On Saturday 09 March 2002 01:59 pm, you wrote:
> On Sat, Mar 09, 2002 at 10:46:46AM -0500, Jeff Darcy wrote:
> > "Seems to" is the key phrase.  It doesn't provide any kind of guarantee,
> > or even improve availability by any determinate amount.  Furthermore, as
> > more people join the race to stay at the top of everyone's cache the
> > benefit to each decreases...while network traffic continues to increase. 
> > At a certain point this becomes tantamount to a deliberate denial of
> > service, and should be treated by the network as such.  I think what some
> > people are suggesting here is something that improves availability
> > deterministically and without increasing network traffic, even if it ends
> > up being less than an ironclad availability guarantee (which might not be
> > possible in this context).
>
> The most promising research direction to increase Freenet's data retention
> (ie. to increase the probability that people find data that was inserted
> into the network at some time in the past) is to look at conditional
> caching.
>
> For example, having a higher caching probability closer to the origin of
> the data along a request-reply path should increase node specialization and
> reduce the rate at which data is dropped.  At the moment, all data is
> cached.
>
> I have been encouraging people to volunteer to experiment with simulations
> to this effect, but unfortunately there haven't been any takers yet.
>
> Ian.

----------------------------------------
Content-Type: application/pgp-signature; charset="us-ascii"; 
name="Attachment: 1"
Content-Transfer-Encoding: 7bit
Content-Description: 
----------------------------------------

_______________________________________________
freenet-tech mailing list
[EMAIL PROTECTED]
http://lists.freenetproject.org/mailman/listinfo/tech

Reply via email to