Re: elevator algorithm bug in ll_rw_blk.c

Matthias Urlichs (smurf@noris.de)
24 Nov 1998 14:59:54 +0100

Riley Williams <rhw@bigfoot.com> writes:
>
> Also, there have been comments about a reverse scan being bad as the
> head may have to reference sectors earlier on the same track. I don't
> buy that, simply because not all disks make use of a 1:1 interleave
> internally,

But you don't know about the boundaries between surfaces, and you do have
milti-block requests which no sane person would want to split up.

That is, if you read sector N+2, then N-1 and N, then N-3, the disk needs
to switch surfaces three times if N-1 happens to be on a different surface
than N.

For disks with embedded servo information (most of them, these days,
because the temperature difference between platters is non-negligible),
switching surfaces is NOT a cheap operation. Therefore, a bidirectional
elevator would slow things down.
The second reason for slowdown is that it would defeat the disk's readahead
algorithm.

NB: Are there disks which can be tuned WRT the way the do caching and/or
readahead? Under Linux, I'd like the disk to do readahead. With MesS-DOS,
caching is likely to be more effective...

-- 
Matthias Urlichs  |  noris network GmbH   |   smurf@noris.de  |  ICQ: 20193661
The quote was selected randomly. Really.    |      http://www.noris.de/~smurf/

-- 
You are faithful to duty, adaptable to environment, loyal to friends.


-
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/