Fault tolerance is a critical capability of hard real-time systems. Even with faults, fault tolerant scheduling algorithms are able to guarantee the real time property of tasks. In primary-backup based fault tolerant scheduling algorithms, only a small time window is left for the backup when the primary faults occur, therefore the backup will likely miss its deadline. This paper proposes a fault tolerant global scheduling with non-preemptive backups (FTGS-NPB). By assigning the highest priority to all backups, the backup can attain processor immediately in case of primary faults, and keep executing until finishing its job. In this way the backup can achieve the shortest response time. The schedulability tests are set up based on deadline analysis and response time analysis. The compatibility of priority assignment algorithms and schedulability tests is discussed. The simulation results show that FTGS-NPB can reduce the amount of additional processors for achieving fault tolerant capability.