On Fri, Apr 04, 2003 at 09:22:26AM -0800, Ian Clarke wrote:
> At my request, Hui Zhang, one of the people who did that paper on an 
> alternative caching scheme, did some research on probabilistic caching 
> (what he called "Inverse-hop Caching").  You can find his results at:
> 
> http://netweb.usc.edu/~huizhang/New_topic/inverse_hop_caching.ps
> 
> I also created a PDF version at:
> 
> http://hawk.freenetproject.org:8080/~ian/inverse_hop_caching.pdf
> 
> Ian.

1. He is not simulating the pcaching scheme we are implementing. We are
implementing P(cache) = 0.8 ^ hopsSinceReset.
He is simulating P(cache) = hopsSinceReset ^ k; optimum value for k
seems to be 1<k<2. It would be interesting to see the two compared.

2. A network of 900 nodes is a bit small, although the inserts to
requests ratio looks reasonable.

3. Where he says "load", he is talking about the amount of data we are
trying to cache. The other paper says the number of hops to find data is
the same either way. But it should improve reachability. The paper
points out that "enhanced clustering" gets much better results; reading
parts of the previous paper, "enhanced clustering" is a scheme where the
node will cache content according to its proximity to the seed key,
which would seem to be unsuitable for a network with freenet's goals
because it would probably introduce many new attacks, and reduce the
network's resilience to attacks against a single document. Also, the
number of keys per node in the simulations reaches quite ridiculous
levels - we cannot reasonably expect good reachability for keys on a
network with 20 keys generated per node where each node caches 40 keys!

> 
> -- 
> Ian Clarke                                                [EMAIL PROTECTED]
> Coordinator, The Freenet Project            http://freenetproject.org/
> Founder, Locutus                                      http://locut.us/
> Personal Homepage                                 http://locut.us/ian/



-- 
Matthew Toseland
[EMAIL PROTECTED]/[EMAIL PROTECTED]
Full time freenet hacker.
http://freenetproject.org/
Freenet Distribution Node (temporary) at 
http://80-192-4-36.cable.ubr09.na.blueyonder.co.uk:8889/kFAXSlo8MmY/
ICTHUS.

Attachment: pgp00000.pgp
Description: PGP signature

Reply via email to