One of the glories of Solaris is that it is so very observable. Then
there are the many excellent blog posts, wiki entries, and books - some
or which are authored by contributors to this very thread - explaining
how Solaris works. But these virtues are also a snare to some, and it is
not uncommon for even experienced practitioners to become fixated by
this or that.
Aubrey, a 70:30 user to system ratio is pretty respectable. Running at
100% is not so pretty (e.g. I would be surprised if you DIDN'T see many
involuntary context switches (icsw) in such a scenario). Esteemed
experts have assured you that ZFS lock contention is not you main
concern. I would run with that.
You said it was a stress test. That raises many questions for me. I am
much more interested in how systems perform in the real world. In my
experience, many of the issues we find in production are not the ones we
found in the lab. Indeed, I would argue that too many systems are tuned
to run simplistic benchmarks instead of real workloads.
However, I don't think it's helpful to continue this discussion here.
There are some esteemed experienced practitioners "out there" who would
be only too happy to provide holistic systems performance tuning and
health check services to you on a commercial basis (I hear that some may
even accept PayPal).
zfs-discuss mailing list