On Tue, Apr 18, 2000 at 10:36:20AM -0700, Chuck McManis wrote:
Track 'n' on all platters formed one cylinder.
Given that disks had a hard
time seeking quickly, things that you could put on one cylinder could be
accessed like a drum without any head movement. In BSD they added "cylinder
groups" which were collections of cylinders (usually 3, sometimes 5) where
the heads need only move in our out one or two steps from the "home"
cylinder to get to a particular bit of data. Then by clustering data into
cylinder groups you could acheive higher disk throughput then if you put
things onto any old track.
Somebody once told me that the definition of a "real pro" database admin
is somebody who can put the right tables on the right cylinders to
optimize throughput. But I didn't know it was an OS-level feature...
guess it would have to be.
--
_______ Shawn T. Rutledge / KB7PWD ecloud(a)bigfoot.com
(_ | |_)
http://www.bigfoot.com/~ecloud kb7pwd(a)kb7pwd.ampr.org
__) | | \________________________________________________________________
Get money for spare CPU cycles at
http://www.ProcessTree.com/?sponsor=5903