INN-base feeder

[bill davidsen]

In article <20030127183620.GB4910 at wonderland.linux.it>,
Marco d'Itri <md at Linux.IT> wrote:
| 
| On Jan 27, bill davidsen <davidsen at tmr.com> wrote:
| 
|  >I assume that I keep the same files in RAM you do, just not explicitly;-)
| Instead of doing ugly games with a ramdisk and cp you can easily be sure
| the files are going to stay in core by using contrib/mlockfile.c.

I just let the Linux caching keep things in memory, and I believe the
original poster uses a makedbz at boot time, I saw no references to a
"master copy" of the data, and really no need for one.

mlockfile doesn't work over boots, and when I tried it needed to be
stopped and rerun as part of expire. It gave no measurable improvement,
so I assume that the caching is doing a good job keeping things in
memory without intervention. The expire complexity didn't seem worth the
effort, based on actual benchmarks. It might be different on a small
memory machine, clearly.
-- 
bill davidsen <davidsen at tmr.com>
  CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.