IBM will soon cease to sell Solaris 10 on x64 machines.
Oracle, it seems, has decided that companies who have their own Unix iron can't peddle Solaris on their x64 tin. Oracle yanked Hewlett-Packard's Solaris OEM contract, which allows it to distribute Solaris 10 and sell support contracts for it on its ProLiant servers, and now, Solaris is being pulled off IBM's System x rack and BladeCenter blade servers.
Maybe IBM is pulling it, and maybe Oracle is pulling it. Oracle, as usual, refused to comment on the matter, and a human being at Big Blue was not able to shed any light on the situation as El Reg went to press. But what IBM has told customers in an announcement letter is that on August 27 it will stop selling Solaris 10 subscriptions on the relatively few x64-based machines it was certified to run on.
IBM sold standard and premium support contracts for Solaris 10, and when the initial OEM deal was announced in August 2007, Big Blue supported Solaris 10 on its System x3650 (two-socket Xeon), x3755 (four-socket Opteron), and x3850 (four-socket and higher Xeon MP) rack-mounted servers. The operating system was certified on IBM's BladeCenter two-socket HS21 (Xeon-based) and four-socket LS41 (Opteron-based) blade servers too.
IBM has subsequently certified Solaris 10 on this-year's HS22 and HS22V (optimized for virtualization) two-socket Xeon 5600 servers announced this year. Big Blue has not yet announced more modern Opteron 4100 or 6100 blades. IBM has not spiked its Opteron lineup as Oracle has done, but it is sure taking its time getting new Opteron gear in the field.
That said, in the past three years, IBM did expand support for Solaris 10 to its LS21, LS22, and LS42 Opteron-based blades in recent years. Similarly, Solaris 10 has been certified on Big Blue's second-generation of x3850, x3950, x3650 machines announced in early 2009 and the new System x3650 M3 that came out in March of this year.
This doesn't sound like a vendor that was enthusiastic about Solaris 10, to be sure, since this was only a small part of IBM's x64 lineup. But it also doesn't sound like a company that was getting ready to kill off selling support contracts either. In some cases, the only way IBM is going to get a hardware sale at a Sun shop – after pushed Power-AIX, Power-Linux, and then x64-Linux combos and failing – is to sell Solaris on System x or BladeCenter boxes. A hardware sale is better than no sale at all, particularly when Sun was giving IBM a piece of the Solaris action. If I had to guess – and I have to because Oracle and IBM are not talking – I would say that Oracle yanked IBM's Solaris contract.
In the meantime, Oracle's brochure for Solaris 10, espousing "support for a wide range of Oracle and third-party servers, workstations, laptops, and peripheral devices," needs to be updated. So does this sentence: "Oracle provides the industry’s broadest line of x86 systems, powered by AMD Opteron or Intel Xeon processors."
As I said back in June when HP confirmed that Oracle had yanked HP's Solaris 10 OEM contract – after Sun spent years and years to get HP to support Solaris on its ProLiant rack and BladeSystem blade servers – I think Oracle doesn't want anyone with a Unix of its own to be able to have the stop-gap position of being able to peddle Solaris on their machines against Oracle. If Oracle ever wants one thing, it is complete control over a situation.
And that means one of two things: One, Dell's Solaris 10 OEM contract is next. Or, two, Dell will eventually become the preferred x64 server provider for Oracle in North America when Oracle finally gets tired of pretending it actually cares about the x64 server business, perhaps giving Fujitsu a similar preferred status for customers in Europe and Asia/Pacific. (This may take a year or two. It may already be in the works). It would not be surprising to me at all for Dell and Fujitsu to get contracts to manufacture Oracle servers, much as Hitachi makes IBM's low-end mainframes and NEC and Dell makes Unisys boxes.
The "Polaris" port of OpenSolaris to IBM's Power architecture, which was hinted at but which never was given monetary or technical backing from either Sun or IBM, never got off the ground, and IBM's commitment to the "Sirius" port of OpenSolaris to the mainframe was tepid at best – you could not legally load OpenSolaris onto z/VM partitions and low-cost Integrated Facility for Linux (IFL) mainframe engines until November 2008.
As El Reg previously reported, Sun had one person dedicated to the mainframe Solaris port, and that person got the axe in an Oracle layoff in the wake of the acquisition this spring. The code for supporting System z mainframes is in OpenSolaris – it hasn't been deleted – but that is a far cry from having a commercially supported Solaris 10. ®