Paulo Gaspar wrote:
> What I mean is that the cache could accumulate quite some wrong cost
> measurements for long periods of time if they would happen with
> resources having a long "cache life".
Correct, but there is no such thing as a perfect cache. Keep this in
mind. We are dealing with stocastical properties and what we can do is
'converge' toward the optimal solution for the entire system. Single
resources will not be perfectly cached, but the entire system load will
(if there is enough statistics available).
So, please, try to see the entire system and not the single page.
[...]
> What I say is that this kind of error can happen a lot. That will
> have a different cost depending on the characteristics of the system
> (e.g.: longer cache lives => higher cost).
Again, correct. In my paper I clearly wrote that the algorithms
described work best on systems that exhibit time-local behavior, that
is, don't exhibit high frequencies of cost variations.
Like any retroactive control system, it cannot possibly have the same
behavior for all frequencies. The control system I described has a
'low-pass' behavior: works best on slowly moving costs... the problem is
that I couldn't come up with a way to tune the system for higher
frequencies without creating an impossible overhead on the system :/
But if you have a suggestion, I'm more than happy to follow it.
> What is a resource? Is it "Invoice detail view" or is it the "detail
> view of invoice number 5678665"?
A "resource", in this case, is every document fragment at all levels,
the output of each pipeline component.
--
Stefano Mazzocchi One must still have chaos in oneself to be
able to give birth to a dancing star.
<[EMAIL PROTECTED]> Friedrich Nietzsche
--------------------------------------------------------------------
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, email: [EMAIL PROTECTED]
