When untarring pkgsrc, expiring news, and trying other disk-intensive
activities, I'm finding my 1.4.2 and 1.5B (586/K6-2) systems spamming
the disk scheduling, leaving interactive users waiting on a 'hung'
machine for up to a minute.

systat reports disk activity at ~100% and transfers are almost double
the random-seek (bonnie) rate-- presumably the scheduler or whatever
is prioritising over adjacent cylinders, so that other 'random' seeks
are falling into a starvation condition.

Any suggestions for a fix?

-- 
Chris Baird,, <cjb@brushtail.apana.org.au>