Processing on the Amoeba distributed operating system is not fault-tolerant. The only concern of its processing service is to perform load balancing on the existing processors, trying to find the processor that best suits a particular process in terms of memory availability and CPU speed. In this paper we introduce a fault-tolerant processing service for the Amoeba system. This service is used in the implementation of the Seljuk-Amoeba operating environment, which offers processing service with different controlled failure semantics on a perprocess basis.