[rescue] Oracle kills Solaris

[Jonathan Patschke]

On Wed, 6 Sep 2017, Dave McGuire wrote:

>  I lurked on that list until I got sick of it.  I got the strong
> impression that there are a bunch of x86 fanboys on there that think
> anything that's not an x86 PC is "legacy", and they had zero interest in
> supporting SPARC.

You can thank Jonathan Schwartz for that.

I personally decommissioned the last SPARC workstation at my day job
because Sun quit making workstations.  At the time, we'd rather have run
Cadence on Solaris.  It worked, and the engineers had gotten used to it,
but we'd already been through a disruptive recall on our Blade 1500s, and
at the end of our hardware support contract, Sun had literally nothing
better[1] to sell us.

They got replaced with mediocre Dells running Red Hat.  After the
engineers got over redoing a decade of accumulated customization, they
stopped caring.  Dell won.  The engineers were laying out ICs again, and I
had one fewer platform to support.

The least expensive SPARC machine I can find on Oracle's web site is, at
$11k, almost twice as expensive as the least expensive POWER8 system from
IBM and about five times as expensive as the least expensive Proliant from
HPE, all at the same memory size and the same or greater CPU core count.

The clear message from Sun and Oracle is that SPARC is not for you unless
you're doing very wide HPC or whatever silly "throughput computing" niche
the SPARC T was cut-out for.

>  They seemed to fall under what I've been calling the "AS/400 problem",
> people thinking that because there were AS/400s in 1988 that all AS/400s
> are from 1988.  Just because there were SPARC systems in 1993 doesn't
> mean all SPARC systems are from 1993.

No, but all SPARC workstations are at least a decade old, and it's been
nearly that long since Sun, Oracle, or Fujitsu competed at
price/performance in small systems.  If they cared about open-source
support for their platform, they'd at least release full documentation and
have some sort of hobbyist buy/lease program.  No one is going to pay $11k
to port a fork of an abandoned proprietary OS to undocumented hardware.

When the only samples of a platform at reasonable affordability are
end-of-life, that's a workable definition of "legacy."


[1] The Ultra 25 did have a less infuriating chassis, I suppose.
-- 
Jonathan Patschke
Austin, TX
USA