On Fri, Jun 21, 2002 at 02:39:10PM +0100, David Laight wrote:

 > Actually it is [1], and is partially responsible for the delays!
 > The problem is that a given request can be delayed for a very
 > long time because of a sequential read/write to a different part
 > of the disk.
 > 
 > 	David
 > 
 > [1] I suspect the 'escalator' algorithm is actually used.

Ah, "escalator" being "seek up, reset to beginning, seek up again"?

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>