On Wed, Jul 30, 2003 at 02:27:40AM -0700, Rudi Cilibrasi wrote: > I do believe the results are significant in that they are not just a > lucky run or fluke; I tried many different parameter combinations and > have seen the SVM perform better in most cases.
I wasn't so much interested in the statistical significance of the performance difference between the algorithms - so much as whether their predictions were actually proving useful (for either algorithm) > My > suspicion is that the simple version of routing that I have heard about > involving forwarding to "closest keys" may be creating some congestion > bottlenecks or file "masking" effects, analogous to how a hash table using > linear probing will create more clustered distributions than a hash > table using a secondary hash function. I am not sure I understand what you mean here - these clustered distributions in hashtables are caused by key-collisions which, given that we are using 160 bit hashes - will never realistically happen in Freenet. Keys do cluster in Freenet, we call this specialization, but it is a desirable effect - and indicates that a node is well integrated into the network. > I think it's clear that there will > be some sort of "self-organizing complexity" as Wolfram would phrase it, > but am still trying to decide how we might characterize and measure > different large-scale network behaviors when under the control of different > routing schemes. Well, there have been extensive simulations of the current "closest keys" scheme. In terms of trying NGrouting - we can make some useful measurements locally - such as the distribution of incoming requests - and the average difference between estimated and actual routing times. > Perhaps it will be useful to add some optional > request-tracking to the protocol so that we can look at actual traces of > a query flowing through the Freenet, and see where the actual holdup > occurs, usually. About a year ago we did something like this - there is a parameter you can set called "watchme" in the config which will cause your node to upload message data to a central database. You can see a graph of an actual request here: http://freenetproject.org/ss/index.php?page=7 We haven't used this functionality in quite a while - although we could reinstate it. For security, nodes with their "watchme" parameter set were forbidden from communicating with non-watchme nodes, meaning that they formed a completely separate network. This raised the question of how representative the watchme network was of the real network, but it was the best we could realistically do. I think the more immediate test will be just to let NGR loose in the network and monitor locally collected data to evaluate its performance relative to the current scheme. Ian. -- Ian Clarke [EMAIL PROTECTED] Coordinator, The Freenet Project http://freenetproject.org/ Founder, Locutus http://locut.us/ Creator, WhittleBit http://whittlebit.com/ Personal Homepage http://locut.us/ian/
pgp00000.pgp
Description: PGP signature
