Re: page faults

Mohammed Hamed wrote:

> Well yes that's what I have been doing! but using the delta rule
> shows no big improvement on the values and they still show a ratio
> around 0.45 and it's still increasing by time

Hmm.. I did some measurements on Linux 2.0.34 and it seems that all disk
reading counts as major page faults even if it is from the disk buffer.
This makes page faults quite useless as a performance measure on Linux.

Lets hope 2.2 has a more generic buffer management approach that does
not differentiate between reads and writes in such a stupid way as Linux
2.0 does.

> which shows that free never dropped below (420*4)=1680

Fine. Your machine does not block on memory. Pagefaults is not a problem
on your machine.

---
Henrik Nordstrom
Spare time Squid hacker