System level policies for fault tolerance issues in the FERMI project

The FERMI system, performing acquisition and DSP of calorimeter data in high energy collision experiments, planned at the LHC collider (CERN, Geneva, CH) is briefly overviewed. The system relies mainly upon the FERMI module, a dedicated VLSI multichip device performing most of the above functions, which is to be installed in large quantities (around 10/sup 5/) in the immediate neighborhood of the collider itself, requiring rad-hard features. The issues for a system which absolutely requires fault diagnosis and possibly fault tolerance are described, with regard to the FERMI module itself.