Asynchronous File I/O on Linux

[bogstad-e+AXbWqSrlAAvxtiuMwx3w at public.gmane.org (Bill Bogstad)]

[replying to my own note]

On Wed, May 19, 2010 at 4:18 PM, Bill Bogstad <bogstad-e+AXbWqSrlAAvxtiuMwx3w at public.gmane.org> wrote:

> patterns where this is good enough. ? My interest in this thread is
> "What mechanisms are available to applications that don't have access
> patterns that fit well with the heuristics that developers are willing
> to put into the kernel?" ? pread(), readahead(), AIO all seem relevant
> to this question.

posix_fadvise(fd, offset, len, advice) lets you tell the system what
the access pattern on an FD is going to be.  You can also
pre-cache regions as well.

Turns out that posix_fadvise(..., POSIX_FADV_WILLNEED) is basically
functionally equivalent to the Linux readahead() call.

Except that my Linux documentation says that readahead() is blocking
while posix_fadvise(...,POSIX_FADV_WILLNEED) is not.  So
posix_fadvise()  would seem to be much friendlier to single-threaded
applications.

Other flags include:

# define POSIX_FADV_NORMAL	0 /* No further special treatment.  */
# define POSIX_FADV_RANDOM	1 /* Expect random page references.  */
# define POSIX_FADV_SEQUENTIAL	2 /* Expect sequential page references.	 */
# define POSIX_FADV_WILLNEED	3 /* Will need these pages.  */
# define POSIX_FADV_DONTNEED	4 /* Don't need these pages.  */
# define POSIX_FADV_NOREUSE	5 /* Data will be accessed once.  */

Looks like great stuff for high IO computing.

Bill Bogstad