Tandem Computers was founded in 1974 by James “Jimmy” Treybig after Hewlett Packard rejected his vision for a truly fault-tolerant computer. The company developed the “NonStop” line of machines, which were engineered to remain operational even in the event of hardware failure, software crashes, or human error. These systems became the invisible backbone of the global financial infrastructure, eventually handling the vast majority of the world’s ATM transactions and stock market trades. Despite its technical success and a legendary employee-centric culture, Tandem eventually struggled against the industry shift toward commodity hardware and standardized operating systems (like UNIX). In 1997, Tandem was acquired by Compaq, and in 2002, through Compaq’s merger with HP, the company returned to the fold of Hewlett Packard.
Historical Context
- The Reliability Crisis (Early 1970s): Business computers of the era were single-point-of-failure systems; a hardware failure often meant total downtime, with some reports suggesting typical mean time between failures was as low as one week.
- The Rise of NonStop (1974-1980s): Founded in Cupertino, Tandem filled a critical gap for banks and stock exchanges that could not afford any downtime. The first system was delivered to Citibank in 1976.
- The Commoditization Era (1990s): The market shifted from specialized, high-cost fault-tolerant hardware to cheaper, interchangeable parts and standard software (Unix/Intel), which squeezed Tandem’s margins.
- The Full Circle (1997-2002): Tandem was bought by Compaq to give them a “credible” installed base in banking. When Compaq merged with HP in 2002, Tandem effectively returned to its origin.
Facts and Persons
- James “Jimmy” Treybig: The visionary founder and CEO known for his rumpled appearance and for implementing a people-first corporate culture.
- Founding Team: Mike Green (Software), Jim Katzman (Hardware), and Jack Losdow-No (Finance).
- Kleiner Perkins: The venture capital firm (Eugene Kleiner and Tom Perkins) that took a massive risk on Tandem in its early days.
- Jim Gray: A database scientist who provided the statistical proof that fault tolerance worked, showing the gap between NonStop systems (years between failures) and conventional systems (weeks).
- David Mackie: The salesman who created the “Mackie Diagram,” a visual guide to the system’s redundancy that every employee carried on their badge.
Technical and Scientific Discoveries
- Shared-Nothing Architecture: Unlike conventional multi-processors that fight over a single shared memory, Tandem processors had their own memory and communicated via a high-speed “Dynabus.” This allowed for nearly linear scalability.
- Fault Tolerance & Redundancy: Every critical path (power, disk, processor, bus) was duplicated. If one failed, another immediately took over without stopping the system.
- “Fail Fast” Philosophy: A design priority to detect a fault and shut down the faulty component immediately to prevent the “silent corruption” of bank records.
- Process Pairs: Ensuring reliability by running two copies of a program on separate processors, with the backup maintaining checkpoints to take over if the primary failed.
- ServerNet: An internal interconnect technology developed by Tandem that later evolved into the industry standard Infiniband.
Consequences
- Financial Infrastructure: Tandem’s success enabled the rollout of the modern ATM network and the stability of global stock exchanges.
- Influence on Software: The “Wolfpack” project with Microsoft brought Tandem’s clustering and high-availability concepts to Windows NT servers.
- Cultural Legacy: Tandem’s “HP Way” inspired culture—including sabbaticals, stock options for all, and “Friday beer busts”—became a benchmark for early Silicon Valley entrepreneurship.
- Operational Reality: The ultimate proof of the design was exemplified by a system that remained running even after the cabinet physically fell over during an earthquake.
