Curt Sampson wrote (originally to port-i386):
>Well, it depends on the workload, actually. From what I've read of
>the merged VM/cache systems (and I am, admittedly not the sort of
>expert that John Dyson is on these matters--so if he disagrees,
>he's right :-)) huge amounts of linear disk I/O can start chucking
>lesser used processes out of memory, when the lesser used processes
>will run again and the disk I/O won't happen again.
>
>I've heard from people who've seen this sort of thing actually
>happen on SunOS machines.
>
>cjs
>
>Curt Sampson    cjs@portal.ca           Info at http://www.portal.ca/
>Internet Portal Services, Inc.
>Vancouver, BC   (604) 257-9400          De gustibus, aut bene aut nihil.

Most of my experience has been with HP's MPE/iX OS. It's not just
lesser-used/priority processes that are affected.  Anything
that needs pages which have not been accessed "recently" gets hosed.  They
wind up waiting for pages to come in, which gives the hog a chance to flush
even more.

What are the preferred solutions to this kind of problem?  (Besides
buying more hardware :)



---
Mike South                       msouth@ablecom.net
Santa Clara, CA, USA