On Friday, Jul 18, 2003, at 08:16 America/Guayaquil, Andreas Hochsteger wrote:
Stefano Mazzocchi wrote:Exactly! I came to that same exact conclusion from this very point: I
tried to come up with an optimal way to know which object to throw aaway
when the cache is full, and I found out that we were heuristically
trying to estimate which one was saving more resources... but without
even trying to measure it and thinking that time is the only valuable
resource.
Throwing away a resource should be determined by the same cost function and not only by 'time' as the following example shows:
There is a resource R1 in the cache, which consumes a quite high ammount
of cache memory and many other resource R2..Rx with low cache memory
consumption, where each takes slightly less production time then R1.
If you only take time as the measurement, then you have to purge some of
R2..Rx, since their cache efficiency is little less than R1.
If you look at the global cache efficiency, you're likely see that removing just R1 might be a better solution, since you can use the free space to cache much more other resources.
I don't know, if your RT suggests such a behaviour (although I've read it completely, but I forget fast these days ;-), I wanted to express it in my own less academic (an perhaps less accourate) words.
Thanks. You are right on target.
--
Stefano, happy to see the academic language barrier diminishing and the info coming across :-)
