On Tue, Nov 04, 2003 at 12:14:33PM +0000, Roger Hayter wrote:Another comment: I have been playing with David <someone>'s latency test. Assume it works as described, and inserts random keys from one node and immediately retrieves them from another, the results are really interesting. Time to retrieve varies from 2 seconds to >8hours (it crashes when the node is restarted so I can't say how much >). There is no obvious median, I usually get a few between 10 seconds and 5-10 minutes. This suggests routing has major problems in that some parts of the key space are largely inaccessible from one of my nodes. I shall try the nodes the other way round, and wonder if David would like to get his latency test to output the keys of his random files so we can see what sort of pattern in the key space they form (i.e., broad regions of keyspace inaccessible, or is it variable within one narrow region?).
Is there any correlation with file size? Are these all the same size?
With the proviso that I am not a programmer, and know nothing about Python (except it has a silly name), a quick glance suggests it inserts KSKs of about 20 bytes of text, and the same size within a byte or two. We should really ask David McNab (who is probably impressed with my innocent trust in running it without knowing what it does). It would be nice if it inserted about 500 random keys at once and did a graph of their position in keyspace against retrieval time (logarithmic!), and superimposed subsequent runs on the results. Latest run is 35s, 2s, 2s, 56s, 709s, 744s, 3172s, 18898s (not yet found).
--
Roger Hayter
_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
