Fwd: Re: fs/etx2/file.c question from postgres developers...

hi...

as per the question of linux getting in the way of reads/writes, i forwarded
bruce's (?) message re: the ext2 code in the linux kernel... this is what i got
back.. hope this helps....

----------  Forwarded Message  ----------
Subject: Re: fs/etx2/file.c question from postgres developers...
Date: Tue, 2 Nov 1999 14:40:51 +0000 (GMT)
From: "Stephen C. Tweedie" <sct@redhat.com>


Hi,

On Mon, 1 Nov 1999 12:12:12 -0700, "Aaron J. Seigo" <aaron@gtv.ca> said:

> in response to a performance issue with disk i/o when using the
> postgresql database (maintained and developed mostly by bsd users,
> though used by many linux users) the following has come up...

> ____________________________________

>> Next question is what to do about it.  I don't suppose we have any way
>> of turning off the OS' read-ahead algorithm :-(.  

> Look what I found. I downloaded Linux kernel source for 2.2.0, and
> started looking for the word 'ahead' in the file system files.  I found
> that read-ahead seems to be controlled by f_reada, and look where I
> found it being turned off?  Seems like any seek turns off read-ahead on
> Linux.

It's a lot more complex than that --- check the do_generic_file_read()
code in mm/filemap.c for the full algorithm.  The readahead "window" is
tuned dynamically based on sequential accesses, and any read outside the
window clears the readahead.

--Stephen
-------------------------------------------------------

-- 
Aaron J. Seigo
Sys Admin