parallelization and the storage manager
br+inn at panix.com
Tue Mar 26 19:16:07 UTC 2002
On Sun, Mar 24, 2002 at 12:42:43AM +0900, Katsuhiro Kondou wrote:
> In article <20020322170635.GA9882 at panix.com>,
> Ben Rosengart <br+inn at panix.com> wrote;
> } They seem to be entirely I/O-bound. It hurts me to look at my ten
> } disks full of news and see that one is at 100% utilization while the
> } others, and the CPU, are almost completely idle.
> How about disks for overview? I suspect that'd be a
> bottleneck rather than article spool when rebuilding.
I have overview striped across eight disks, and each disk also
contains a cycbuff. The pattern I see is that one disk is maxed
out for a while, then the next, then the next, in the same order
that they're mentioned in the metacycbuff definition in cycbuff.conf.
So for me, at least, overview is not the bottleneck. I'm using ovdb.
Ben Rosengart (212) 741-4400 x215
1. A robot may not injure entertainment industry profits, or, through inaction,
allow entertainment industry profits to come to harm. --Matt McLeod
More information about the inn-workers