Was SPARC done in by Solaris? The Linux-on-SPARC experiment was perhaps a very recent one, for the last year or so. And of course, it was abandoned almost as soon as it was started. But would it have made a difference to SPARC if the Linux-on-SPARC plan had been kicked off say, 5 years ago?
8 replies (most recent on top)
Linux on SPARC is something that was tried a lot of times before Exadata SL over the years. There were a measurable number of people in SPARC hw -- more often VSP than ESP, but some even there -- who were sure Solaris was what was holding their sales back. Linux on SPARC never, ever took hold, other than among some tinkerers. (See also: Linux on POWER)
The flip side: lots of Sun customers during the depths of SPARC performance woes took their Solaris workloads and happily ran them on x86. These are customers Sun would have completely lost if Solaris x86 hadn't been there for them.
This would have been even more successful if there'd been more support for that on the HW side, and if Sun didn't do things like, oh, what was the word -- "defer" Solaris x86 -- right when it would have been a major weapon to retain dissatisfied SPARC customers. Instead we got people in VSP who said -- "gee, look, no one ever picks Solaris preconfigure on x86! They all pick Windows or Linux!"
To quote Scott: well, duh.
Windows and Linux had a much bigger overall (that is, not just Sun) market share than Solaris on low end x86 servers-- no one ever argued otherwise. I personally was shocked whenever the numbers for preconfig matched reality. (For the sarcasm impaired: I was never actually shocked by this.)
Not to mention that stats showed that very few customers used Solaris preconfig, even on SPARC. Solaris configurations are deeply personal choices :-).
There were lots of things Sun could have done to make x86 much more successful (see "how about not killing it and then unkilling it when it became obvious how stupid that was", above ). They actually started doing most of those things in the early 2000s. Oracle undid it all, as fast as they could, immediately after the acquisition.
Oracle DB was running fine on Linux on SPARC, hence the Exadata SL6 product. Let's not talk about the performance compared to running on Solaris/SPARC though.
As for the OPs question; I'd say their fates were heavily intertwined. Without SPARC, Solaris has no significant enough x86 market share and would lose out to Linux on that platform "because everyone knows/runs it". Conversely, SPARC without Solaris, well there's just no market for it, the main market was legacy customers that were on Solaris/SPARC for many many years.
Linux on sparc was LE's idea and he pulled out after MH promised opc is the way to go. They pulled out people from solaris and sparc engineered systems team to build Linux on sparc. Finally they killed engineered systems using sparc and the sparc cloud effort was dead on arrival.
We can sell sparc to interested third party instead let people go
But now that Solaris is gone, wouldn't Oracle have to port the DB, middleware etc to Linux after all?
None of customers wants Sparc on Linux.
The last poster is correct. Solaris vs Linux is like Betamax vs VHS. It's not about superior technology, it's all about adoption.
Linux on SPARC has no business case unless you also have major applications
running on Linux on SPARC. Ie. for Oracle that'd imply porting the entire
database and middleware stack over to Linux on SPARC too.
Just the Linux on SPARC thingy along is nothing then a PoC. without any money to gain.