The legacy of dual boot PA-RISC hardware

HP’s legacy hardware includes both a Unix and an MPE system. Everybody knows about HP 9000s and HP 3000s. Not as many people know this was the same hardware for more than a decade. One story about 9000s and 3000s shows how you can use the same box for each legacy OS.
The accepted wisdom is that a special processor dependent code unit takes HP’s hardware and makes it Unix or MPE. Instead, a former HP support engineer just uses a boot-level command, something executed from a BCH prompt.
For years, processor dependent code on ROM was forcing the HP 3000-styled K-Class systems into MPE/iX boot mode only. Software created by some well-regarded MPE vendors made its way into unscrupulous hands, defeating HP’s passworded utility software. At that moment it results in a way to designate an HP 9000 system as an MPE-bootable server. We might have called it re-flashing the PDC ROM at the time. It’s been a few years.
The password-protected utility was being used by HP’s support engineers. As the passwording was defeated, HP’s iron could be configured any way a customer wanted. Selling much-cheaper K-Class 9000 servers as if they were 3000s became a way to purchase a resold K-Class from a broker and save tens of thousands of dollars, and in some cases even more. It led to an HP lawsuit against the rogue brokers like Hardware House (the worst offender). There were jail sentences handed down to two other brokers (house arrests) while one of the Hardware House owners turned in state’s evidence in exchange for dropped charges.
Quite the cause celeb, the move seemed to show the MPE customer base that HP still recognized the inherent value in its MPE-related intellectual property. The lawsuits and HP’s High Tech Crimes Taskforce rose up in 1999 and 2000. It was a time when Y2K remediations and rewrites gave the 3000 some cover in the war over the datacenter and business computing. An HP business decision not two years later made the battle over the MPE IP moot, though.
Once the A-Class and N-Class servers arrived, a different config program began to be used. It couldn’t be defeated by outside software. HP had also shifted to a processor-linked pricing model for the 3000 and MPE. That meant the outrageous markups for the K-Class 3000s, the regrettable tiered pricing, disappeared for the newest 3000s. To escape the tiered-pricing jail, customers could buy new servers.
It hardly matters the way it once did. The upgrade HP created to PA-RISC, Itanium, is being discontinued by Intel any day now. The rewrite of MPE for Itanium was shut down after an estimate of the cost didn’t pass executive approval. HP 3000s might now number less than 5,000. But knowing that today an engineer with the right know-how can pull an HP server from a packing crate — and boot either legacy OS on it — feels like the magic the 3000 market once needed. An HP 9000 sold for a fraction of its identical hardware counterpart, right up to the end of HP sales of the 3000.
HP’s argument, a good one in concept, was that MPE and Image made the 3000 worth so much more than a 9000. A big problem was that the servers were being sold against one another by the HP sales force. The commercial application lead which MPE once had over HP-UX was gone. The pricing disadvantage HP put these 3000s at did its part to drag down growth of the line.
That support engineer, or any former HP technical worker, has nothing to do with HP’s regrettable decision to kill off its MPE business. That’s a business decision based on a forecast of an ecosystem that HP controlled with its alliances, marketing, and engineering designs. At one point in the 3000’s history, the inability to buy raw K-Class hardware and designate it either MPE or HP-UX mattered, though. It’s a delight to learn how the engineering was supposed to work.

Leave a Reply