|
Hello Wensong,
I wonder why it is done in such a complicated way? I know most of it is
It isn't complicated. :)
Ok.
from the net/sched/estimator.c but why not using the direct way of
computing the average?
Something like following should work:
avg = ((time - timeframe) * avg + new) / time
Your formula is the same as avgrate = avgrate*(1-W) + rate*W, where W is
timeframe/time. Oh, you miss something in your formula, it should be
avg = ((time - timeframe) * avg + timeframe * new) / time
Yes, precisely. Thanks for the hint.
Why are the numbers scaled by 2^5?
For more precise numbers, we keep the 1/32 fraction numbers in the
counter.
Right, no FP in KS! :)
Would it make sense to use the ip_vs_estimator to write a scheduler
which chooses the RS with either the least amount of bps or the least
amount of pps?
Yes, it's possible, but it will make ipvsadm a little bit more complicated
to set those bps/pps thresholds.
It isn't complicated. :)
Best regards,
Roberto Nibali, ratz
--
echo '[q]sa[ln0=aln256%Pln256/snlbx]sb3135071790101768542287578439snlbxq' | dc
|
