Matthew Toseland wrote: > Local requests obviously use far less *output* bandwidth than remote > requests.
Interesting - any idea why? This seems to suggest that there's a lot of backtracking at the bottom of the search tree. If the average degree is about 20 and there are a few thousand nodes, it should be possible to reach any node in 3 hops. Maybe the current HTL is excessive, causing inserts and unsuccessful requests to backtrack into irrelevant parts of the network? Cheers, Michael
