MASSCOMP

~ the obit ~

The Massachusetts Computer Corporation came out of an idea belonging to C. Forbes Dewey at MIT and an inventor named Chester Schuler, and was formally founded in Westford in 1981. Everyone called it MASSCOMP. It went after a market that the general purpose computer companies treated as a nuisance: laboratories and test rigs and anything else where the computer has to keep up with the physical world rather than with a person typing.

The thing it did that nobody else had productized was continuous high rate acquisition. The MC-500 could sample an analog signal a million times a second and write every sample to disk indefinitely, on Motorola 68000 hardware running at a few megahertz and disks that were slow even by the standards of 1983. Getting a megasample a second off a converter is one problem. Never dropping one while the operating system is also doing everything else an operating system does is a different and much worse problem, and the answer was RTU, a real-time Unix that MASSCOMP wrote itself.

For a company that size it built an unreasonable amount of its own software, including its own C and Fortran compilers and its own high level graphics subsystem. The odd one was Laboratory Workbench, which let you assemble a real-time system as a block diagram, wiring analog inputs into signal processing blocks and out to instruments drawn on the screen without writing any of the plumbing. That is a visual programming environment for data acquisition, shipped in the middle of the 1980s. The business worked. Sales reached around 65 million dollars a year.

Then it made a bet that ended it. On 1 August 1988 a merger was announced with Concurrent Computer Corporation, a real-time computer company from Monmouth County, New Jersey that had been spun out of Perkin-Elmer in 1985 and traced back to Interdata before that. Concurrent had 2,800 employees against MASSCOMP's 700 and had done 247 million dollars of revenue in 1987. MASSCOMP was the buyer. It paid 241 million dollars, about two thirds of it borrowed on junk bonds, which also bought out Perkin-Elmer's majority stake. The transaction closed on 27 September 1988. The surviving legal entity was MASSCOMP. The surviving name was Concurrent.

~ the rap sheet ~

~ a megasample a second ~

The technical problem MASSCOMP solved is the one I work on for a living, which is the only real connection I have to a company that was retired when I was four. A sensor produces samples at a fixed rate that does not care what your software is doing, and every part of the path from the converter to storage has to be bounded, or you lose data and cannot get it back. On a drone today that means a camera feeding a small computer that has a hard budget on latency. In 1983 it meant a laboratory instrument and a 68000 and a disk, and the reason a whole company could be built on it is that the general purpose Unix of the period could not make any timing promises at all.

That is why RTU mattered enough to survive the merger. The MASSCOMP flavor of real-time Unix became the operating system on Concurrent's Series 5000 and Series 6000 machines, so the engineering outlived the brand by years. The part that got expensive is now a scheduling policy you can turn on in any kernel.

~ the minnow and the whale ~

Buying a company three times your size and then adopting its name along with its New Jersey headquarters is a strange shape for a takeover, and it was unusual even in an era full of leveraged deals. James K. Sims had been running Concurrent and kept running the merged company. Production moved from Westford to Concurrent's New Jersey plant. On paper the logic was clean, with MASSCOMP's lower end real-time products and its Unix expertise sitting underneath Concurrent's larger machines at a moment when Unix was clearly becoming the thing customers wanted.

In practice the debt did not go away. The early 1990s recession arrived almost immediately, there were rounds of layoffs in Monmouth County, the two engineering organizations fought over which products to build, and IBM and Digital turned out to be difficult to undercut in real-time. Sims was out by 1990 and Denis R. Brown replaced him. Another change came in 1993 with John Stihl. Revenue peaked around 300 million dollars in 1990 and was down to 179 million by 1994. An industry analyst asked about it later said the merger did not produce anything for Concurrent except problems.

The debt was finally worked down from about 200 million dollars to under 25 million by 1995, seven years after the deal. By then MASSCOMP had been gone longer than it had existed.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »