IBM authorizes OpenSolaris on mainframes
Step one comes after step two
A month ago, Sine Nomine Associates, the mainframe consultancy that has done most of the porting work on the "Sirius" variant of Solaris Unix to IBM's mainframe platform, said that the code was available for people to try out on the OpenSolaris project site. And now IBM has come around to actually authorize the use of OpenSolaris on its z/VM partitioning software for mainframes.
In an announcement letter to customers, IBM said that as of November 18, it was perfectly okay with mainframe shops using its Integrated Facility for Linux - a combination of a special variant of z/VM partitioning software and lower-priced "specialty" mainframe engines - to run OpenSolaris, the open source distribution of the Solaris operating system created by Sun. The IFL engines are said to cost roughly a quarter the price of a real mainframe engine, and the absorption of some Linux workloads by mainframe shops has been one reason that the IBM mainframe market has not collapsed since 2000. (IBM also has other specialty engines, called the zAAP, for running Java, and the zIIP, for accelerating DB2 database routines, and together with the IFL, these three engines have saved the mainframe biz.)
Prior to this announcement, the IFL licensing terms only allowed customers to use Linux on these engines and, equally importantly, only allowed Linux to interact with the z/VM operating system (which is akin to a hypervisor on a X64 box as well as an operating system supporting applications in its own right) and applications running atop z/VM. With the change in licensing terms, which applies to every IFL that IBM has shipped on any zSeries or System z mainframe since introducing them in 2000, when IBM decided that its entire server line would be made to run Linux. These days, about a quarter of the raw mainframe capacity that IBM sells (as measured in MIPS) is supporting Linux workloads.
While it is highly unlikely that Solaris will equal Linux in terms of customer enthusiasm and revenues for Big Blue, every little bit helps, and by putting Solaris on equal footing (in terms of IFL and z/VM licensing, not technically), the company has officially opened the door to the possibility of customers supporting some Solaris work on mainframes. That said, there is not a policy of trying to absorb Solaris onto the mainframe in an aggressive manner, as was the case with the mainframe, as far as I can ascertain from my conversations and attempted clarifications from the marketeers and techies at IBM. (The executives I have talked to in the Unix and mainframe parts of the IBM house have basically passed the buck on the question so far.)
As I pointed out a month ago, IBM and Sun announced an OEM agreement in August 2007, which saw IBM embrace Solaris and go so far as to sell support contracts for Solaris on selected models of its System x rack and BladeCenter blade servers. At that time, Bill Zeitler, the top exec in charge of the Systems and Technology Group, gave a nod of sorts to the mainframe and Power ports, but neither Sun nor IBM have made any big moves to get Polaris, the Power port of OpenSolaris, and Sirius, the mainframe variant, out there as a commercially supported variant of Solaris 10. Zeitler retired on August 1, and it is unclear what commitment, if any, Bob Moffatt, the executive who has replaced him, has to Solaris on either Power or mainframe systems.
With the change in authorization, mainframe shops are allowed to run Linux or OpenSolaris on z/VM partitions atop the IFL engines, and OpenSolaris can reach out into z/VM facilities and applications as well, just like Linux. IBM cautions that it is not authorizing the use of other operating systems on the IFLs.
The Sirius distro of OpenSolaris is available in binary code form through Sine Nomine's website and the open source code is available through the OpenSolaris site, which is where development for OpenSolaris for X64 and Sparc processors as well as the related "Polaris" project for Power-based machines is hosted. ®
Sponsored: Benefits from the lessons learned in HPC