Re: Page Cache question

Re: Page Cache question

[Robert Hancock]

Adnan Khaleel wrote:
> So would it be possible to me to disable it for just a particular user space process?

It's the core of the whole VM/memory management subsystem. It's not 
something you can just disable.

It would be better if you explained what exactly you were trying to 
accomplish..

> 
> Adnan
> 
> 
>> I'm looking for a way to disable the page cache for an
> 
>> experimental NUMA system running the 2.6.17 kernel. I would prefer to
> 
>> only disable the page cache for my process and still have it be enabled
> 
>> by the rest of the system. Is there an easy way of doing this?
> 
>> Alternatively, I would be fine disabling the Page Cache altogether as well.
> 
> 
> 
> All user memory is in the page cache. Without it you could only run kernel
> 
> modules.
> 
> 
> 
> -Andi

-- 
Robert Hancock      Saskatoon, SK, Canada
To email, remove "nospam" from hancockr@nospamshaw.ca
Home Page: http://www.roberthancock.com/

Re: Page Cache Question

[Eugene Teo]

<quote sender="Helge Hafting">
> Adnan Khaleel wrote:
>> I'm looking for a way to disable the page cache for an
>> experimental NUMA system running the 2.6.17 kernel. I would prefer to
>> only disable the page cache for my process and still have it be enabled
>> by the rest of the system. Is there an easy way of doing this?
>> Alternatively, I would be fine disabling the Page Cache altogether as 
>> well.
>>   
> Assuming what you really mean is that you don't want to cache
> file i/o for that process - try opening files with O_DIRECT.

You will probably find this tool useful:
http://userweb.kernel.org/~akpm/pagecache-management/

"... a little tool which permits the management of the pagecache usage of
arbitrary applications.  Effectively it prevents the targetted application
from using any pagecache at all."

Eugene

Re: Page Cache Question

[Helge Hafting]

Adnan Khaleel wrote:
> I'm looking for a way to disable the page cache for an
> experimental NUMA system running the 2.6.17 kernel. I would prefer to
> only disable the page cache for my process and still have it be enabled
> by the rest of the system. Is there an easy way of doing this?
> Alternatively, I would be fine disabling the Page Cache altogether as well.
>   
Assuming what you really mean is that you don't want to cache
file i/o for that process - try opening files with O_DIRECT.

Helge Hafting

Re: Page Cache Question

[Andi Kleen]

Adnan Khaleel <akhaleel2001@yahoo.com> writes:

> I'm looking for a way to disable the page cache for an
> experimental NUMA system running the 2.6.17 kernel. I would prefer to
> only disable the page cache for my process and still have it be enabled
> by the rest of the system. Is there an easy way of doing this?
> Alternatively, I would be fine disabling the Page Cache altogether as well.

All user memory is in the page cache. Without it you could only run kernel
modules.

-Andi

Re: Page Cache Question

[Tino Keitel]

On Sun, Aug 05, 2007 at 09:32:59 -0700, Adnan Khaleel wrote:
> 
> I'm looking for a way to disable the page cache for an
> experimental NUMA system running the 2.6.17 kernel. I would prefer to
> only disable the page cache for my process and still have it be enabled
> by the rest of the system. Is there an easy way of doing this?
> Alternatively, I would be fine disabling the Page Cache altogether as well.

Maybe this helps for you:

http://lwn.net/Articles/224653/

Regards,
Tino

Re: Page Cache question

[Adnan Khaleel]

So would it be possible to me to disable it for just a particular user space process?

Adnan


> I'm looking for a way to disable the page cache for an

> experimental NUMA system running the 2.6.17 kernel. I would prefer to

> only disable the page cache for my process and still have it be enabled

> by the rest of the system. Is there an easy way of doing this?

> Alternatively, I would be fine disabling the Page Cache altogether as well.



All user memory is in the page cache. Without it you could only run kernel

modules.



-Andi

-





       
____________________________________________________________________________________
Building a website is a piece of cake. Yahoo! Small Business gives you all the tools to get online.
http://smallbusiness.yahoo.com/webhosting 

Page Cache Question

[Adnan Khaleel]

I'm looking for a way to disable the page cache for an
experimental NUMA system running the 2.6.17 kernel. I would prefer to
only disable the page cache for my process and still have it be enabled
by the rest of the system. Is there an easy way of doing this?
Alternatively, I would be fine disabling the Page Cache altogether as well.



Thanks



AK






       
____________________________________________________________________________________
Pinpoint customers who are looking for what you sell. 
http://searchmarketing.yahoo.com/

Re: page cache question

[Nick Piggin]

Xin Zhao wrote:
> Sorry if this question is dumb.
> 
> Linux uses address_space to identify pages in the page cache. An
> address space is often associated with a memory object such as inode.
> That seems to associate the cached page with that inode. My question
> is: if a file is closed and the inode is destroyed, will the cached
> page be removed from page cache immediately?  If so, does that mean

Yes. The inode's struct address_space contains the radix tree which
indexes the pagecache pages.

> the file system has to load data from disk again if a user promptly
> open and read the same file again? If not, how does linux determine
> when to evict a cached page? using LRU?
> 

Yes they would have to be read again. However in general the inode is
not destroyed after the file is closed -- inodes are cached too.

-- 
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 

page cache question

[Xin Zhao]

Sorry if this question is dumb.

Linux uses address_space to identify pages in the page cache. An
address space is often associated with a memory object such as inode.
That seems to associate the cached page with that inode. My question
is: if a file is closed and the inode is destroyed, will the cached
page be removed from page cache immediately?  If so, does that mean
the file system has to load data from disk again if a user promptly
open and read the same file again? If not, how does linux determine
when to evict a cached page? using LRU?

Thanks in advance for your kind help!

-x