Abstract: In distributed environments, runtime failures often occur. If the distributed system has the ability to handle such failures dynamically (within runtime), it is said to be fault tolerant. Such systems suffer from the problem of being slow if compared to other non-fault tolerant systems. Moreover, if the system is based on a Distributed Shared Memory (DSM) in exchanging data among the distributed application members, then it is going to be slower and may be inefficient. In this study, a generic DSM based Fault Tolerance Protocol (FaTP) is introduced. FaTP is a high performance fault tolerance protocol. The proposed protocol is based on the Linda Tuple space DSM model. It introduces a compact set of DSM access primitives and supplied with a fault tolerance layer based on dynamic replication. The complexity of FaTP has been measured and its performance has been evaluated.