> On Mon, 26 Jul 1999, Gerald Aigner wrote:
>
> > While optimizing the program I discovered that disk I/O
> > performance in Linux could be improved substantially. In
> > particular, there are two main problems:
> >
> > 1) The kernel copies data many times before it ends up in user
> > space.
>
>
> I'm very sorry, but you seem to be too late. This problem is
> (at least partially) fixed in the newer 2.3 kernels...
>
I just tried the 2.3.11 kernel. The new kernel is slightly faster
(67 seconds instead of 69 seconds) and it uses less CPU time (37% vs. 45%)
However 37% is still a very high CPU utilization.
> > 2) Using mmap doesn't help because it serializes I/O within a single
> > process (you can't read simultaneously from two disks).
>
> A single process uses a single thread of control, and yes,
> we do readahead over RAID0 sets...
>
Using mmap on the linux2.3.11 kernel takes now 105 seconds, which is
still slower than doing read system calls in separate threads.
However overall CPU utilization is lower (around 7%) using mmap.
So is there an easy way to make mmap as fast as read (in terms of
wall clock time).
-Gerald
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/