Hello,
as fas as I can see from fdatasync man page, and from the latest kernel
sources (2.4.1ac3, fs/buffer.c), they are equivalent.
Using of fdatasync in database can gain significant gain on systems which
supports it (on HP it gains up to 25% with pg_bench on PostgreSQL 7.1b5).
Are there any plans to implement this correctly? And due to what problems it
was not implemented yet?
--
Sincerely Yours,
Denis Perchine
----------------------------------
E-Mail: [email protected]
HomePage: http://www.perchine.com/dyp/
FidoNet: 2:5000/120.5
----------------------------------
On Sunday 18 February 2001 18:22, Denis Perchine wrote:
> Hello,
>
> as fas as I can see from fdatasync man page, and from the latest kernel
> sources (2.4.1ac3, fs/buffer.c), they are equivalent.
>
> Using of fdatasync in database can gain significant gain on systems which
> supports it (on HP it gains up to 25% with pg_bench on PostgreSQL 7.1b5).
>
> Are there any plans to implement this correctly? And due to what problems
> it was not implemented yet?
Forget this crap. Seems I missed these lines:
err = file->f_op->fsync(file, dentry, 0);
err = file->f_op->fsync(file, dentry, 1);
--
Sincerely Yours,
Denis Perchine
----------------------------------
E-Mail: [email protected]
HomePage: http://www.perchine.com/dyp/
FidoNet: 2:5000/120.5
----------------------------------