See #10.
As I said — years and years of work to become what Solaris is.
Look at the threading, scheduling, resource management inside Solaris for SMP machines with 8 or more CPU’s. Sun sells servers up to 64 CPU’s and 256GB of core.
Everyone thinks it is easy to make an operating system.
It is today - for one CPU. For two isn’t easy stuff to master and do well - it took Linux/*BSD several years after dual-CPU PC’s became common to get to where SunOS/Solaris was in 1994 or so.
Getting to the point of supporting 64 CPU’s, with scalable results? That’s still way down the road for Linux/*BSD.
Getting the availability/reliability/failure containment features of Solaris in any Linux-like distro? Even more years away.
There are Linux advocates out there, trying to show that Oracle on Linux is faster for some big number of users or big number of queries. Typically, these benchmarks fail in that they’re not a true apples:apples comparison, often due to differences in the storage architectures (eg, SCSI vs. SATA, or local SATA vs. network attached storage, etc) and so on. In some of them, I’ve seen them talk about having a power glitch or some other problem that takes down the RDMBS or whole machine — and when they come back up, the Oracle running on Solaris is A-OK, and the Linux implementations have drives out of sync, database transactions missing, etc. And the Linux advocates just wave these results off.
In a database environment, the integrity of the data is paramount - speed is second.
For Oracle as a RDBMS company, they’re dependent upon these features existing for their high-end DBMS customers, so if Sun and Solaris just faded away into the sunset, Oracle would be screwed.