Re: [PATCH 1/2] aio: add vectored I/O support

[Avi Kivity]

Joel Becker wrote:

> On Thu, Oct 14, 2004 at 01:10:01PM -0700, Yasushi Saito wrote:
>  
>
> > This is a patch against 2.6.9-rc3-mm3 that add supports for vectored 
> > async I/O.  It adds two additional commands, IO_CMD_PREADV and 
> > IO_CMD_PWRITEV to libaio.h. The below is roughly what I did:
> >    
>
> 	How does this differ substantially from lio_listio() of each I/O
> range?  Does it have some significant performance win, or is it just
> aiming for a completeness that POSIX doesn't (to my knowledge) specify?
>
>  
It is a huge performance win, at least on the 2.4-based RHEL kernel. 
Large reads (~256K) using 4K iocbs are very slow on a large RAID, while 
after I coded a similar patch I got a substantial speedup.

I don't know the cause for the slowdown; maybe request merging only 
works for queued requests, and as the RAID has a large TCQ depth, 
requests didn't have much of a chance to queue in the kernel.


--
Do not meddle in the internals of kernels, for they are subtle and quick to panic.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/