Emulation reclaims its time
January 2, 2009
It's been seven years since HP announced there would be no faster 3000s. Early in the transition era the homesteading advocates in the community pumped up the ideal of an emulator, hardware that would make up for the 3000s which HP would be stripping out of its product lineup. At the time the talk served little more purpose than to give the homesteaders a cause to rally around. The new generation of 3000s was twice as fast as the Series 900 predecessors, fast enough for a good while.
But more than 80 months have passed and computing power requirements have rolled upward. The market learned that the newer generation of 3000s was better connected and faster, but few in number. HP's late delivery of the N-Class and A-Class hampered production. If you needed a faster 3000 than the top-end 900 Series, you hunted for N-Class servers that customers were returning once they migrated.
Now the climate and demand has changed for faster 3000 compute power, but there's no relief from HP on the horizon. Staying with MPE/iX solutions means a customer needs to keep planning for more connectivity and speed. An emulator that can leverage the latest Intel chip designs, rather than flog the familiar PA-RISC architectures of HP, might find a market by next year.
Why next year, rather than, say, 2004? The used 3000s of five years ago ran fast enough to replace MPE/iX systems and justify the investment. Now only a rare, 4-way N-Class offers that kind of power leap. And there's nothing built upon PA-RISC that can network and integrate like an Intel-based server. The irony of that reality is not lost on the 3000 customer, who saw the Intel/HP generation of 3000 first promised, then denied to the community.
Back in the first years of transition the OpenMPE meetings revolved around assembling an emulator coalition. Nobody could make a market out of selling such a product, the community reasoned, if they had to compete with one another.
In time, however, grabbing control of MPE/iX and its prospects overtook emulation on the OpenMPE issues list. Source code licensing was on the list of six key needs for a homesteading community. The challenge was unlike emulation in one significant way: HP held all the cards for the MPE/iX project. PA-RISC, however, was once a hardware standard promoted to the community. HP called the group the Precision RISC Organization and tried to get major hardware makers to adopt the architecture. As a consequence, the specifics of PA-RISC internals are not the top secret that MPE/iX has always been.
Emulator vendors need MPE/iX expertise to make a product of any use to the 3000 market. That's why the likes of Allegro Consultants has been among many discussions of emulators. The consultants there are the hands-down most experienced in PA-RISC, and some have written modules of MPE/iX.
In recent months, the deeper thinkers of the 3000 community have kicked around what a 3000-on-Intel MPE/iX would need to deliver the 3000's business advantages. Roy Brown, a developer of more than 30 years on the platform, said this in October:
The paradigm here must be the PA-RISC machine’s HP 3000 Compatibility Mode (even if HP can’t spell it properly), which let users of Classic systems port them, at the binary level and with the absolute minimum of change, often none, to PA-RISC.
For an MPEmulator to attract existing HP 3000 business users, it must aspire to that level of compatibility, and even if the underlying processes handle things quite differently, the goal must be that every invocation of an openly documented MPE process is accepted, and produces effectively the same results as it would have on a real HP3000.
An emulator vendor can take advantage of what's been built to emulate PA-RISC to make a two-step 3000 emulation product. First, the product must boot up a Linux distro to demonstrate its capabilities. Then a seasoned MPE/iX development team must test the hardware design with MPE/iX tasks to give it HP 3000 capabilities.
The rising tide of value in a 2010 emulator will be interconnectivity. HP will take no more steps to link advanced storage devices and networking protocols to HP 3000s. We're heard no reports of anybody using the SCSI Pass Through driver, one of the last lab projects finished for the 3000, to connect bigger and faster drives to the 3000. But a honking-big Intel-based server which already has those connections ready can deliver more than horsepower to a 3000 homesteader. That kind of emulator delivers the future. With the economy in a stall and migration plans stretched out in some places, selling an emulator to extend a 3000's life looks like a better offer than it did seven years ago.