The problem of adequate yield in the production of VLSI devices is first considered. The authors then point out that, in the near future, a complete computing system on a single chip is to be expected. The number of elements required is so large that unavoidable production and other faults and errors will make yields unacceptably low.
The solution proposed is to introduce redundant elements that can be used to replace faulty ones either statically or dynamically; that is, either after initial testing or automatically during use.
Probabilistic equations are derived that allow efficiency estimates to be made for different models, and these are used to indicate numerical trends. This is a well written and valuable contribution to the literature.