I received an email in my work inbox this morning from Oracle, subject: If It Must Work, It Runs on Solaris.

I agree.

It’s also too late.

Solaris x86 was always the red-headed stepchild at Sun. It had a poor HCL, it lagged behind for patches, very little work was done on optimizing compilers for it, and in turn, it was poorly supported by vendors too, despite (theoretically) porting to it being naught but a recompile. Oracle in particular considered it very much to be “tier 2” in terms of support, patches and upgrades. And why would you run a mission-critical system on something that not even its owner takes seriously? Sun were obviously afraid of cannibalizing their SPARC sales, and that’s fair enough, but seriously, they were bold enough to give away Solaris source code…

If large organizations running SPARC across the board had been offered the opportunity to move to Solx86 (x64 now) on their non-critical systems and it would “just work”, that would have been a no-brainer. And it’s not as if this was technically insurmountable; back in the days of NeXT you could compile a fat binary that would run on any of its hardware platforms. Apple moved its users from 68k → PowerPC → Intel. DEC moved its customers from VAX to AXP, and how HP has OpenVMS on Itanic. Good luck getting a Linux binary to work seamlessly even on the same hardware.

Instead it was easier to move to a whole new OS, and even Oracle jumped on that bandwagon. People say that Linux will save organizations money but it is more subtle than that. Adding a new OS – whatever it is – to an existing estate will only increase cost. Linux can only save you money if you use it to replace two or more existing platforms, on the OS side, or if you use it to get onto cheaper commodity hardware (which implies replacing at least one current OS). Sun’s x64 kit, like the x4500 (aka Thumper) and x4600, was great, clearly ahead in storage and compute density. A unified Solaris running on x86, x64 and SPARC is a compelling offering, from the desktop to the datacentre. Or would have been.

It’s really unclear to me, as a long-term Oracle professional, where exactly they’re going with this, and when planning mission-critical infrastructure, uncertainty is the killer.