Re: stochastic fair queueing in the elevator [Re: [BENCHMARK] 2.

[Andrea Arcangeli <andrea@suse.de>]

On Mon, Feb 10, 2003 at 03:49:59PM +0100, Giuliano Pochini wrote:
> 
> >> You can wait 10 minutes and still such command can't grow.  This is why
> >> claiming anticipatory scheduling can decrease the need for readahead
> >> doesn't make much sense to me, there are important things you just can't
> >> achieve by only waiting.
> >> 
> > 
> > The anticipatory scheduler can easily permit 512k of reading before seeking
> > away to another file.  In fact it can allow much more, without requiring that
> > readhead be cranked up.
> 
> IMHO anticipatory scheduling and readahead address different problems. RA is
> simpler and cheaper. Reading a few more KB comes almost for free and that
> helps a lot sequential reads. AS is useful for random i/o (fs metadata,
> executables, ...), but it wastes time if the timer expires, or if the new
> request wants data which is far away the previous one. AS is useful for

Yep.

> sequential reads too, but only if the application reads large chunks of
> data, otherwise RA is better. I we need both RA and AS to address the

actually even if the app reads large chunks of data RA is needed, the
size of the read/io_submit syscalls won't reach the
readpage/wait_on_page at the pagecache layer, w/o readahead we would
read PAGE_SIZE per dma and not more (ignoring PAGE_CACHE_SIZE of course,
which isn't going to be enough anyways due mem fragmentation issues)

> largest variety of workloads.
> 
> Bye.


Andrea