Beautiful Architecture

Posterity

From 1990 on, a number of factors contributed to a decline in Tandem’s sales:

• Computer hardware in general was becoming more reliable, which narrowed Tandem’s edge.

• Computer hardware was becoming much faster, highlighting some of the basic performance limitations of the architecture.

In the 1990s, the T/16 processor architecture was replaced by a MIPS-based solution, though much of the remaining architecture remained in place. On the other hand, the difference in performance was big enough that as late as 2000, Tandem was still using the MIPS processors to emulate the T/16 instructions. One of the reasons was that most Tandem system-level software was still written in TAL, which was closely coupled to the T/16 architecture. Moves to migrate the codebase to C were rejected because of the cost involved.

For such a revolutionary system, the Tandem/16 has made a surprisingly small impression on the industry and design of modern machines. Much of the functionality is now more readily available—mirrored disks, network file systems, the client-server model, or hot-pluggable hardware—but it’s difficult to see anything that suggests that they happened by following Tandem’s lead. This may be because the T/16 was so different from most systems, and of course the purely commercial environment in which it was developed didn’t help either.