It looks like, especially from Hal's logs, that we are seeing a lot of
restarting of chains when it isn't necessary. My experimenting last night did
not seem to produce too much of this (I sent a message for an unexistant key
with htl 100 which TimedOut after about a minute with no such messages), but I
suppose as the number of dead connections in the datastores increase, so does
the time taken for each hop.
The first thought might just be to decide that 3 seconds was too optimistic,
and start using 5 for the default. Another thought is that the problem is
greater for the low HTL towards the end, where the problems of one node is not
balanced out by the fact that for a node that has no problems it should not
take all of 3 seconds - and that one might add a constant term to equation
(3*HTL + 10 or something).
An approach that is probably more rewarding, is to be statistical about it.
Statistics says that with probability 90%,
time < HTL*(expected time per hop) + 1.28 * sqrt(HTL) * (standard deviation of
time per hop)
(given that it is normal distributed, which it won't be - but statistics
always assumes everything is normal distributed in my experience)
If one assumes a expected value of 2 seconds, and a deviation of 4, one gets
something like:
3*HTL 90% confidence
3.0000 7.1200
6.0000 11.2408
9.0000 14.8681
12.0000 18.2400
15.0000 21.4487
18.0000 24.5414
21.0000 27.5462
24.0000 30.4815
27.0000 33.3600
30.0000 36.1909
...
60.0000 62.8973
...
120.0000 112.3817
...
300.0000 251.2000
So for a standard request with a HTL of 20 or so, the first node would wait
about as much time as it does now, while the ones towards the end wait quite a
bit longer then they do today.
What do you people think? Is 90% confidence to low? What is a good estimate at
the expected value and variation of the time per hop (without good estimates
here, the confidence means nothing)?
--
Oskar Sandberg
md98-osa at nada.kth.se
#!/bin/perl -sp0777i<X+d*lMLa^*lN%0]dsXx++lMlN/dsM0<j]dsj
$/=unpack('H*',$_);$_=`echo 16dio\U$k"SK$/SM$n\EsN0p[lN*1
lK[d2%Sa2/d0$^Ixp"|dc`;s/\W//g;$_=pack('H*',/((..)*)$/)
_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev
