Nineteen springtimes ago, HP was offering an operating system to run alongside MPE on the same hardware. To say that HP's Multiple Operating System Technology was virtualization might be an overstatement. But the unreleased product gave Unix and MPE equal footing in a single hardware system. MPE was the cradle that Unix would rest in, much like Linux is the cradle where the PA-RISC virtualization rests in the Stromasys Charon product. The only reason it was not released might have been the horsepower demands on the hardware. MOST was not starved off the price list by a lack of HP desire from the 3000 division. But the daring of its engineering was on a battleground between HP's own products.
I worked on external communications for MOST for Hewlett-Packard in the spring of 1995. It was one of the biggest assignments I took on during the months that led up to creating the 3000 NewsWire. The audacity of putting a venerated OS in as a bootstrap system for HP-UX apps led me to believe HP was exploring every prospect to win any customer who was veering toward the market's magnetic pull of Unix.
HP showed off external specifications for MOST to key partners in '95. The product was scheduled to emerge in the fall of that year on Series 9x9 and 99X PA-RISC systems. These were the highest horsepower 3000s in the HP stable. MOST was to begin with two partitions, one for MPE/iX and the other for HP-UX. Or, a customer could run two separate instances of MPE on a single server. MPE was to be the primary partition, controlling the uptime of the hardware.
In one sense, this product wouldn't have been a 3000 -- because half of it would be dedicated to running Unix apps and processes. Independence, a white paper on the product stated, "is especially important, as the co-dependencies between the different OS should be as small as possible."
MOST might have been ahead of its time in hardware requirements, but it reminds me of the virtualization that nearly every operating system enjoys today. The Stromasys Charon lineup, the VMware partitions which run Windows, Linux, and Mac OS all at once -- all of these flow from the concept that drove MOST. Well, there's a major difference. HP didn't release MOST, even after a beta test period and surveys that showed most of the customers saw it as an evolutionary path to heterogenous computing.
"The future path is almost impossible to foresee," HP's briefing stated. "Windows or OS/2? WARP? Unix or NT? Once proprietary, but now open systems?"
The software would have realized the founding principle of PA-RISC engineering: "Eventually, any PA-RISC operating system will be able to operate concurrently and independently on the same hardware platform."
HP delivered on some of these promises many years later, employing its Superdome designs for high-end servers with flexible partitions. This was not strictly emulation, because the native hardware remained the same. It's a sad piece of history that by the time Superdome was rolled into the markets, MPE/iX was not an environment supported on the high-priced server.
The OS came closest to its rightful place as keystone of HP's business computing strategy with MOST, however. HP said that it "is a natural complement to the four strategic directions of the HP 3000:
- Reinforce HP 3000's strengths in mission critical OLTP environments
- Superior integration in a multi-platform environment
- Provide an evolution to client/sever computing
- Deliver innovative applications and services
The Hewlett-Packard of 1995 was looking for a way to "let customers add, test and develop new applications without purchasing a new Unix box." That might have been the downfall for MOST. A successful server, steered by MPE but also able to run Unix apps, would surely have been a roadblock to more HP 9000 server sales. HP bet hard on Unix in that era, a play that now seems to have run out of step with the Windows and Linux choices of today.