On 8/28/2018 2:42 AM, arnold(a)skeeve.com wrote:
But I can testify from experience at my last job that
AIX, HP/UX and
Solaris are all still in heavy use at major companies, mainly hosting
big database systems, but still there and not going away too soon.
If you haven't tried Solaris 11, at least 11.2 or later, check it out. A
lot of GNU stuff is available, and you can "pkg install" apache, tomcat,
PHP, MYSQL, and a slew of other things that are actually being kept up
to date. The recent PHP 7.1.17 exploit fix was put out in the regular
Solaris 11 SRU a few weeks after it was publicly released. And that was
supposedly after the "mass layoff" of Solaris engineers. I've also
noticed quite a few bug fixes for Solaris internals and device drivers
since then so Oracle is still maintaining Solaris. I think their most
recent support time line puts Solaris support out to 2030 or so.
I've installed and maintain a few Solaris 11.1 (since upgraded to 11.3)
clusters for Oracle databases on Intel (Dell) blades with fiber channel,
and with ZFS they've been rock-solid.
Side note, and I probably already said this a long time ago here, but
back in the early days of Linux kernel 2.6, I complained on one of the
mailing lists about the removal of a way to control the size of the disk
cache, along with the tendency for the kernel to page out applications
in favor of more disk cache. The snobby answer I got back from a
developer was that (paraphrased) "We know better than you about memory
allocation" or some such garbage. I've always turned my nose up at Linux
since then. I figure if that level of arrogance had infested it to that
degree, and an open-source project at that, I wanted no part of it.
Since then, of course, they have made some changes that make it less
likely to do that, but I have a recent Oracle Linux (Redhat) system here
running Oracle eBusiness that is currently 2GB into swap, and has 10G of
disk cache. Swappiness is set to 0, and other tunables were altered that
should have stopped that. That's just plain dumb.