I can't reboot or unmount in the middle of the query.
option 3 sounds good. its for debugging and benchmarking.
but not just, I am doing only sequential access on some of my
algorithms and OScaching is useless and it seems I can't use huge
relations to defeat the cache replacement
(since the operation I am doing is very long).
I am caching myself inside the algorithm, the OS caching
is just wasting memory and does not know whats best
for my algorithm. 

Is there a file system out there
that I could disable its caching when mounting?
I have seen some unfamiliar ones but I prefer something in
the kernel source.
Can you give me some direction on where I can change the code
to disable caching?
10x.

Regards,
        tzahi.

> -----Original Message-----
> From: guy keren [mailto:[EMAIL PROTECTED] 
> Sent: Friday, April 29, 2005 2:52 AM
> To: Tzahi Fadida
> Cc: [email protected]
> Subject: Re: cleaning memory.
> 
> 
> 
> On Thu, 28 Apr 2005, Tzahi Fadida wrote:
> 
> > I am using the /proc/sys/vm/block_dump
> > to see reading and writing of blocks for a process.
> > I am looking at the postgresql database process.
> > When I run a query for the first time I see a lot of
> > READs. but the next times I run it, it doesn't show
> > READs, suggesting the kernel have cahced those blocks 
> somewhere. Even 
> > if I get out of the program and reenters it doesn't help. I also 
> > disabled the swapfile just in case(probably not related).
> > How can I flush this caching?
> 
> 1. reboot the machine ;)
> 2. umount and re-mount the file system whose files you want un-cached
>    (upon umount, all cached pages of the umounted file system 
> are freed by
>    the kernel).
> 3. modify the kernel's source to avoid caching on a per-filesystem
>    (perhaps add an option to the mount command stating "no cache"?).
>    not sure how easy this is ;)
> 
> 
> something else about caching (summary: the cache holds both 
> data and meta-data, and it is the later that's more expensive 
> to access): 
> http://www.ussg.iu.edu/hypermail/linux/kernel/0209.0/0929.html
> 
> 
> -- 
> guy
> 
> "For world domination - press 1,
>  or dial 0, and please hold, for the creator." -- nob o. dy
> 
> 



=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]

Reply via email to