[Opendnssec-develop] Signer Testplan: first try
Rick van Rein
rick at openfortress.nl
Wed Apr 8 20:09:40 UTC 2009
You woke up the tiny mathematician living inside my head!
> In my world, jitter is ABS(MAX(VARIANCE(signature expiration time))).
1. You must mean MAX(ABS(...)) and not ABS(MAX(...))
2. I don't think you intended to use VARIANCE as a consolidation function,
which is how statistics defines it; I think your intention is to calculate
a difference with a mean value and have the consolidation done by MAX?
3. Instead of VARIANCE (which is a square of the (standard) deviation from
the mean value) you probably mean the deviation.
I think I would capture your intentions in formulae as follows:
M = Average_i (sigexptime_i)
Jitter = Max_i (Abs (sigexptime_i - M))
Average_i and Max_i are consolidation functions ranging over i,
and sigexptime_i is the time of the i'th signature expiration.
> signature expiration = calculated expiration time - jitter +
> (random(jitter) * 2)
> where random(x) is a function generating a random numberr such as 0 ≤
> r ≤ x. this would generate a signature exception that can vary +/-
> some jitter number of seconds, right?
> >- Is random jitter acceptable?
> not only acceptable, it is required.
I missed the background of jitter, perhaps we can clear it up?
Is Jitter a feature (a way of spreading signatures over time if a lot
of work would otherwise come in at once) or a bug (the result of the
non-ideal OS and such)?
I cannot help but feel that we are re-inventing realtime scheduling here.
Hope this helps,
More information about the Opendnssec-develop